Malware Alerts

Subscribe to Malware Alerts feed
Online headquarters of Kaspersky Lab security experts.
Updated: 9 min 23 sec ago

The Chronicles of the Hellsing APT: the Empire Strikes Back

Tue, 04/14/2015 - 22:30

Introduction

One of the most active APT groups in Asia, and especially around the South China Sea area is "Naikon". Naikon plays a key part in our story, but the focus of this report is on another threat actor entirely; one who came to our attention when they hit back at a Naikon attack.

Naikon is known for its custom backdoor, called RARSTONE, which our colleagues at Trend Micro have described in detail. The name Naikon comes from a custom user agent string, "NOKIAN95/WEB", located within the backdoor:

NOKIAN string in Naikon backdoor

The Naikon group is mostly active in countries such as the Philippines, Malaysia, Cambodia, Indonesia, Vietnam, Myanmar, Singapore, and Nepal, hitting a variety of targets in a very opportunistic way. What was perhaps one of the biggest operations of the Naikon group was launched in March 2014, in the wake of the MH370 tragedy that took place on March 8th. By March 11th, the Naikon group was actively hitting most of the nations involved in the search for MH370. The targets were extremely wide-ranging but included institutions with access to information related to the disappearance of MH370, such as:

  • Office of the President
  • Armed Forces
  • Office of the Cabinet Secretary
  • National Security Council(s)
  • Office of the Solicitor General
  • National Intelligence Coordinating Agency
  • Civil Aviation Authority
  • Department of Justice
  • National Police
  • Presidential Management Staff

The Naikon group used mostly spear-phished documents for the attacks, with CVE-2012-0158 exploits that dropped the group's signature backdoor.

While many of these attacks were successful, at least one of the targets didn't seem to like being hit, and instead of opening the documents, decided on a very different course of action.

The empire strikes back

Here's a question - what should you do when you receiving a suspicious document from somebody you don't know, or know very little? Choose one:

  • Open the document
  • Don't open the document
  • Open the document on a Mac (everybody knows Mac's don't get viruses)
  • Open the document in a virtual machine with Linux

Based on our experience, most people would say 2, 3 or 4. Very few would open the document and even fewer would actually decide to test the attacker and verify its story.

But this is exactly what happened when one of the Naikon spear-phishing targets received a suspicious email. Instead of opening the document or choosing to open it on an exotic platform, they decided to check the story with the sender:

Naikon target asks for confirmation of the email

In the email above, we can see the target questioning the authenticity of the Naikon spear-phishing. They ask the sender if it was their intention to email this document.

The attacker was, of course, not confused in the slightest, and being very familiar with the internal structure of the target's government agency, replied claiming that they work for the secretariat division and were instructed to send it by the organization's management:

Naikon attacker replies to the target

The reply is written in poor English and indicates that the attacker is probably not as proficient in the language as the intended victim. Seeing the reply, the target obviously decided not to open the document. Moreover, they decided to go a bit further and try to learn more about the attacker.

Not long after the first exchange, the following email was sent to the attacker by the target:

The attachment is a RAR archive with password, which allows it to safely bypass malware scanners associated with the free email account used by the attackers. Inside the archive we find two decode PDF files and one SCR file:

Much to our surprise, the "SCR" file turned out to be a backdoor prepared especially for the Naikon fraudsters.

The file "Directory of ... Mar 31, 2014.scr" (md5: 198fc1af5cd278091f36645a77c18ffa) drops a blank document containing the error message and a backdoor module (md5: 588f41b1f34b29529bc117346355113f). The backdoor connects to the command server located at philippinenews[.]mooo[.]com.

The backdoor can perform the following actions:

  • download files
  • upload files
  • update itself
  • uninstall itself

We were amazed to see this course of action and decided to investigate the "Empire Strikes Back"-door further; naming the actor "Hellsing" (explained later).

The malware used by the intended victim appears to have the following geographical distribution, according to KSN data:

  • Malaysia – government networks
  • Philippines – government networks
  • Indonesia – government networks
  • USA - diplomatic agencies
  • India (old versions of malware)

In addition, we've observed the targeting of ASEAN-related entities.

Victims of Hellsing attacks

The actor targets its intended victims using spear-phishing emails with archives containing malware, similar to the one it used against the Naikon group. Some of the attachment names we observed include:

  • 2013 Mid-Year IAG Meeting Admin Circular FINAL.7z
  • HSG FOLG ITEMS FOR USE OF NEWLY PROMOTED YNC FEDERICO P AMORADA 798085 PN CLN.zip
  • Home Office Directory as of May 2012.Please find attached here the latest DFA directory and key position officials for your referenece.scr
  • LOI Nr 135-12 re 2nd Quarter.Scr
  • Letter from Paquito Ochoa to Albert Del Rosario,the Current Secretary of Foreign Affairs of the Philippines.7z
  • Letter to SND_Office Call and Visit to Commander, United States Pacific Command (USPACOM) VER 4.0.zip
  • PAF-ACES Fellowship Program.scr
  • RAND Analytic Architecture for Capabilities Based Planning, Mission System Analysis, and Transformation.scr
  • Update Attachments_Interaction of Military Personnel with the President _2012_06_28.rar
  • Update SND Meeting with the President re Hasahasa Shoal Incident.scr
  • Washington DC Directory November 2012-EMBASSY OF THE PHILIPPINES.zip
  • ZPE-791-2012&ZPE-792-2012.rar
  • zpe-791-2012.PDF.scr

We've observed RAR, ZIP and 7ZIP archives in the attacks - the 7ZIP archives with passwords were probably introduced as a way to bypass the recent security features on Gmail, which block password-protected archives with executables inside.

Each backdoor has a command and control server inside as well as a version number and a campaign or victim identifier. Some examples include:

MD5 Date C&C Campaign identifier 2682a1246199a18967c98cb32191230c Mar 31 2014 freebsd.extrimtur[.]com 1.6.1_MOTAC 31b3cc60dbecb653ae972db9e57e14ec Mar 31 2014 freebsd.extrimtur[.]com 1.6.1_MOTAC 4dbfd37fd851daebdae7f009adec3cbd Nov 08 2013 articles.whynotad[.]com 1.5_articles.whynotad.com-nsc 015915bbfcda1b2b884db87262970a11 Feb 19 2014 guaranteed9.strangled[.]net 1.5_guaranteed9-nsc 3a40e0deb14f821516eadaed24301335 Mar 31 2014 hosts.mysaol[.]com 1.6.1_imi;simple 73396bacd33cde4c8cb699bcf11d9f56 Nov 08 2013 web01.crabdance[.]com 1.5_op_laptop 7c0be4e6aee5bc5960baa57c6a93f420 Nov 08 2013 hosts.mysaol[.]com 1.5_MMEA bff9c356e20a49bbcb12547c8d483352 Apr 02 2014 imgs09.homenet[.]org 1.6.1_It c0e85b34697c8561452a149a0b123435 Apr 02 2014 imgs09.homenet[.]org 1.6.1_It f13deac7d2c1a971f98c9365b071db92 Nov 08 2013 hosts.mysaol[.]com 1.5_MMEA f74ccb013edd82b25fd1726b17b670e5 May 12 2014 second.photo-frame[.]com 1.6.2s_Ab

The campaign identifiers could be related to the organizations targeted by the specific builds of this APT. Some possible descriptions for these initials could be:

Artifacts and overlap with other APTs

Interestingly, some of the infrastructure used by the attackers appears to overlap (although around a year apart) with a group tracked internally at Kaspersky Lab as PlayfullDragon (also known as "GREF"); while other aspects of the infrastructure overlap with a group known as Mirage or Vixen Panda.

For instance, one of the PlayfullDragon's Xslcmd backdoors described by our colleagues from FireEye (md5: 6c3be96b65a7db4662ccaae34d6e72cc) beams to cdi.indiadigest[.]in:53. One of the Hellsing samples we analysed (md5: 0cbefd8cd4b9a36c791d926f84f10b7b) connects to the C&C server at webmm[.]indiadigest[.]in. Although the hostname is not the same, the top level domain suggests some kind of connection between the groups. Several other C&C subdomains on "indiadigest[.]in" include:

  • aac.indiadigest[.]in
  • ld.indiadigest[.]in
  • longc.indiadigest[.]in

Another overlap we observed is with an APT known as Cycldek or Goblin Panda. Some of the Hellsing samples we analysed in this operation (e.g. md5: a91c9a2b1bc4020514c6c49c5ff84298) communicate with the server webb[.]huntingtomingalls[.]com, using a protocol specific to the Cycldek backdoors (binup.asp/textup.asp/online.asp).

It appears that the Hellsing developer started with the Cycldek sources and worked together with the operators from other APT groups. Nevertheless, it is sufficiently different to warrant classification as a stand-alone operation.

So, where does the Hellsing name come from? One of the samples we analysed (md5: 036e021e1b7f61cddfd294f791de7ea2) appears to have been compiled in a rush and the attacker forgot to remove the debug information. One can see the project name is Hellsing and the malware is called "msger":

Of course, Hellsing can have many different meanings, including the famous doctor from Bram Stoker's Dracula. However, according to Wikipedia, "Hellsing (ヘルシング Herushingu) is also a Japanese manga series written and illustrated by Kouta Hirano. It first premiered in Young King Ours in 1997 and ended in September 2008".

The Hellsing series chronicles the efforts of the mysterious and secret Hellsing Organization, as it combats vampires, ghouls, and other supernatural foes; which makes it perhaps an appropriate name for our group.

In addition to the Hellsing/msger malware, we've identified a second generation of Trojan samples which appear to be called "xweber" by the attackers:

"Xweber" seems to be the more recent Trojan, taking into account compilation timestamps. All the "msger" samples we have seen appear to have been compiled in 2012. The "Xweber" samples are from 2013 and from 2014, indicating that at some point during 2013 the "msger" malware project was renamed and/or integrated into "Xweber".

During our investigation we've observed the Hellsing APT using both the "Xweber" and "msger" backdoors in their attacks, as well as other tools named "xrat", "clare", "irene" and "xKat".

Other tools

Once the Hellsing attackers compromise a computer, they deploy other tools which can be used for gathering further information about the victim or doing lateral movement. One such tool is "test.exe":

Name test.exe Size 45,568 bytes MD5 14309b52f5a3df8cb0eb5b6dae9ce4da Type Win32 PE i386 executable

This tool is used to gather information and test available proxies. Interestingly, it also contains the Hellsing debug path:

Another attack tool deployed in a victim's environment was a file system driver, named "diskfilter.sys", although internally it claims to be named "xrat.sys". The driver is unsigned and compiled for 32-bit Windows. It was used briefly in 2013, before being abandoned by the attackers, possibly due to Windows 7 driver signing requirements:

Another tool used by the attackers is called "xKat":

Name xkat.exe Size 78,848 bytes MD5 621e4c293313e8638fb8f725c0ae9d0f Type Win32 PE i386 executable

This is a powerful file deletion and process killer which uses a driver (Dbgv.sys) to perform the operations. We've seen it being used by the attackers to kill and delete malware belonging to their competitors.

Some of the debug paths found in the binaries include:

  • e:\Hellsing\release\clare.pdb
  • e:\Hellsing\release\irene\irene.pdb
  • d:\hellsing\sys\irene\objchk_win7_x86\i386\irene.pdb
  • d:\hellsing\sys\xkat\objchk_win7_x86\i386\xKat.pdb
  • d:\Hellsing\release\msger\msger_install.pdb
  • d:\Hellsing\release\msger\msger_server.pdb
  • d:\hellsing\sys\xrat\objchk_win7_x86\i386\xrat.pdb
  • D:\Hellsing\release\exe\exe\test.pdb
Attribution

In general, the attribution of APTs is a very tricky task which is why we prefer to publish technical details and allow others to draw their own conclusions.

The Hellsing-related samples appear to have been compiled around the following times:

Assuming normal work starts at around 9 am, the attacker seems to be most active in a time-zone of GMT+8 or +9, considering a work program of 9/10 am to 6/7pm.

Conclusions

The Hellsing APT group is currently active in the APAC region, hitting targets mainly in the South China Sea area, with a focus on Malaysia, the Philippines and Indonesia. The group has a relatively small footprint compared to massive operations such as "Equation". Smaller groups can have the advantage of being able to stay under the radar for longer periods of time, which is what happened here.

The targeting of the Naikon group by the Hellsing APT is perhaps the most interesting part. In the past, we've seen APT groups accidentally hitting each other while stealing address books from victims and then mass-mailing everyone on each of these lists. But, considering the timing and origin of the attack, the current case seems more likely to be an APT-on-APT attack.

To protect against a Hellsing attack, we recommend that organisations follow basic security best practices:

  • Don't open attachments from people you don't know
  • Beware of password-protected archives which contain SCR or other executable files inside
  • If you are unsure about the attachment, try to open it in a sandbox
  • Make sure you have a modern operating system with all patches installed
  • Update all third party applications such as Microsoft Office, Java, Adobe Flash Player and Adobe Reader

Kaspersky Lab products detect the backdoors used by the Hellsing attacker as: HEUR:Trojan.Win32.Generic, Trojan-Dropper.Win32.Agent.kbuj, Trojan-Dropper.Win32.Agent.kzqq.

Appendix:

Hellsing Indicators of Compromise

Microsoft Security Updates April 2015

Tue, 04/14/2015 - 13:58

Microsoft releases 11 Security Bulletins (MS15-032 through MS15-042) today, addressing a list of over 25 CVE-identified vulnerabilities for April of 2015. Critical vulnerabilities are fixed in Internet Explorer, Microsoft Office, and the network and graphics stacks. Most of the critical remote code execution (RCE) vulnerabilities reside in the IE memory corruption bugs for all versions of Internet Explorer (6-11) and the Microsoft Office use-after-free, however, they appear to all be result of private discoveries.

The Microsoft Office CVE-2015-1649 use-after free is a critical RCE impacting a variety of software and scenarios. The vulnerable code exists across desktop versions Word 2007, 2010, the Word Viewer and Office Compatibility apps, but not Word 2013 or Word for Mac. It's also critical RCE on the server-side in Word Automation Services on Sharepoint 2010 and Microsoft Office Web Apps Server 2010, but not SharePoint 2013 or Web Apps 2013.

As the new Verizon Data Breach 2015 report highlighted today, many exploits currently effective against targets are exploiting vulnerabilities patched long ago. According to their figures, many of the exploited CVE used on compromised hosts were published over a year prior. Microsoft provides Windows Update to easily keep your software updated, and Kaspersky products provide vulnerability scanners to help keep all of your software up-to-date, including Microsoft's. Please patch asap.

From the heap of vulnerabilities and fixes rated "Important", the Hyper-V DoS issue effects the newest Microsoft platform code: Windows 8.1 64-bit and Windows Server 2012 R2 (including the Server Core installation, which is fairly unusual). While the flawed code has not been found to enable EoP on other VMs within the Hyper-V host, attacked Hyper-V systems may lose management of all VMs in the Virtual Machine Manager.

Your Tax Refund with a Data Kidnapping Twist!

Tue, 04/14/2015 - 07:40

Oh, how procrastination gets all of us! April 15th is the U.S. tax deadline and it looks like most of us will be coming down to the wire on declaring our taxes and holding our collective breath in expectation of that sweet, sweet refund. Sadly, our malware writing friends are aware of this and their discipline has proven far superior. Knowing that many are on the lookout for emails from the Internal Revenue Service concerning pending refunds, criminals have crafted some of their own:

The attachment is actually a Trojan-Downloader.MsWord.Agent malware, built by the same group behind the recent LogMeIn malicious campaign described here.

The infection scheme is very similar to the aforementioned, however, the threat actor has moved on from abusing Pastebin entries and has instead hacked a Web server in China to host the instructions script file. This file as well as the download URL are also encoded in Base64 and the resulting payload is actually ransomware.

URLs embedded in the malicious macros leading to a Base64 encoded instructions script file and the payload URL below

Instructions files with the URL to the ransomware payload

The malicious ransomware payload is detected by Kaspersky Anti-Virus as Trojan-Ransom.Win32.Foreign.mfbg

Due to the reliance on the IRS branding, this particular malicious campaign is mostly focused on US citizens and permanent residents of the USA.

Challenging CoinVault – it's time to free those files

Mon, 04/13/2015 - 07:23

Some months ago we wrote a blog post about CoinVault. In that post we explained how we tore the malware apart in order to get to its original code and not the obfuscated one.

So when were contacted recently by the National High Tech Crime Unit (NHTCU) of the Netherlands' police and the Netherlands' National Prosecutors Office, who had obtained a database from a CoinVault command & control server (containing IVs, Keys and private Bitcoin wallets), we were able to put our accumulated insight to good use and accelerate the creation of a decryption tool.

We also created a website and started a communications campaign to notify victims that it might be possible to get their data back without paying.

To build the decryption tool we needed to know the following:

  • Which encryption algorithm was being used?
  • Which block cipher mode was being used?
  • And, most importantly, what malware are dealing with?

There was obviously no time for "hardcore" reverse engineering, so the first thing we did was run the malware sample to see what it was doing. And indeed, just as we thought, it was another CoinVault sample. The next thing we did was open the executable in a decompiler, where we saw that the same obfuscation method was used as described in the post. So CoinVault it is. However, we still didn't know which encryption algorithm and block cipher mode it was using.

But luckily we have a sandbox! The nice thing about the sandbox is that it executes the malware, but also has the ability to trace virtually anything. We can dump files and registry changes but in this case the memory dumps were the most interesting. We knew from the previous CoinVault samples that the malware was using the RijndaelManaged class, so all we had to do was search in the memory dump for this string.

And here it is. We see that it still uses AES, although not the 128-bit block size anymore, but the 256-bit one. Also the block cipher mode has changed from CBC to CFB. This was all the information we needed to write our decryption tool (link to decryption tool).

To see if you can decrypt your files for free, please go to https://noransom.kaspersky.com

Simda's Hide and Seek: Grown-up Games

Mon, 04/13/2015 - 00:30

On 9 April, 2015 Kaspersky Lab was recently involved in the synchronized Simda botnet takedown operation coordinated by INTERPOL Global Complex for Innovation. In this case the investigation was initially started by Microsoft and expanded to involve a larger circle of participants including TrendMicro, the Cyber Defense Institute, officers from the Dutch National High Tech Crime Unit (NHTCU), the FBI, the Police Grand-Ducale Section Nouvelles Technologies in Luxembourg, and the Russian Ministry of the Interior's Cybercrime Department "K" supported by the INTERPOL National Central Bureau in Moscow.

As a result of this takedown of 14 C&C servers in the Netherlands, USA, Luxembourg, Poland and Russia. Preliminary analysis of some of the sinkholed server logs revealed a list of 190 countries affected by the Simda botnet.

Simba character, courtesy of Walt Disney Productions, has nothing to do with Simda botnet

Simda is a mysterious botnet used for cybercriminal purposes, such as the dissemination of potentially unwanted and malicious software. This bot is mysterious because it rarely appears on our KSN radars despite compromising a large number of hosts every day. This is partly due to detection of emulation, security tools and virtual machines. It has a number of methods to detect research sandbox environments with a view to tricking researchers by consuming all CPU resources or notifying the botnet owner about the external IP address of the research network. Another reason is a server-side polymorphism and the limited lifetime of the bots.

Simda is distributed by a number of infected websites that redirect to exploit kits. The bot uses hardcoded IP addresses to notifying the master about various stages of execution process. It downloads and runs additional components from its own update servers and can modify the system hosts file. The latter is quite an interesting technique, even if it seems deceptively obvious at first glance.

Normally malware authors modify host files to tamper with search engine results or blacklist certain security software websites, but the Simda bot adds unexpected records for google-analytics.com and connect.facebook.net to point to malicious IPs.

KL detected the #Simda #bot as Backdoor.Win32.Simda, it affected hundreds thousands victims worldwide

Tweet

Why is that, one might ask? We don't know, but we believe that the answer is connected with Simda's core purpose – the distribution of other malware. This criminal business model opens up the possibility of exclusive malware distribution. This means that the distributors can guarantee that only the client's malware is installed on infected machines. And that becomes the case when Simda interprets a response from the C&C server - it can deactivate itself by preventing the bot to start after next reboot, instantly exiting. This deactivation coincides with the modification of the system hosts file. As a farewell touch, Simda replaces the original hosts file with a new one from its own body.

Now, curious mind may ask: how does it help them? Those domains are no longer used to generate search results, but machines infected by Simda in the past might occasionally continue to send out HTTP requests to malicious servers from time to time, even in when exclusive 3rd-party malware is supposed to have been installed.

We need to remember that these machines were initially infected by an exploit kit using a vulnerability in unpatched software. It's highly likely that 3rd-party malware will be removed over time, but a careless user may never get round to updating vulnerable software.

If all those hosts keep coming back to the malicious servers and asking for web resources such as javascript files, the criminals could use the same exploits to re-infect the machines and sell them all over again – perhaps even 'exclusively' to the original client. This confirms once again – even criminals can't trust criminals.

In this investigation Microsoft and various law enforcement bodies completed the sinkholing process and Kaspersky Lab willingly contributed to the preparations for the takedown. That work included technical analysis of malware, collecting infection statistics, advising on botnet takedown strategy and consulting our INTERPOL partners.

Kaspersky Lab detected the Simda bot as Backdoor.Win32.Simda and according to our estimations based on KSN statistics and telemetry from our partners it affected hundreds thousands victims worldwide.

Simda is automatically generated on demand and this is confirmed by the absence of any order in compilation link times. Below is a chart generated from a small subset of about 70 random Simda samples:

Samples link times in UTC timezone

The increase in link times is most likely related to the activity of the majority of Simda victims located somewhere between UTC-9 and UTC-5 timezones, which includes United States.

Thanks to the sinkhole operation and data sharing between partners we have put up a page where you can check if your IP has connected to Simda C&C servers in the past. If you suspect your computer was compromised you can use one of our free or trial solutions to scan your whole hard drive or install Kaspersky Internet Security for long-term protection.

Kaspersky Lab products currently detect hundreds of thousands of modifications of the Simda together with many different 3rd-party malware distributed during the Simda campaign.

Darwin Nuke

Fri, 04/10/2015 - 07:00

In December 2014 we discovered a very interesting vulnerability in the Darwin kernel, which is an open source part of Apple's two operating systems: OS X and iOS. As a result, OS X 10.10 and iOS 8 are also at risk. This vulnerability is connected with the processing of an IP packet that has a specific size and invalid IP options. As a result, remote attackers can cause DoS (denial of service) of a device with OS X 10.10 or iOS 8 installed. It means that attackers can send just one incorrect network packet to the victim and the victim's system will crash.

OS X 10.10 crash after invalid network packet processing

Using vulnerability in the Darwin kernel attackers can cause DoS of a device with OS X 10.10 or iOS 8 installed

Tweet

While analyzing this vulnerability we've discovered that the following devices with 64-bit processors and iOS 8 installed are affected by this threat:

  • iPhone 5s and later models
  • iPad Air and later models
  • iPad mini 2 and later models

To understand the nature of this bug let's look at a crash dump:

Kernel stack trace

You can see from this trace that something went wrong in the icmp_error() function and it calls the panic function. This function tries to construct a new ICMP error message and resend it. This screenshot shows that the icmp_error was called after parsing packet options. The problem lies in this piece of code:

The cause of the problem

When the conditions laid down in the code are met, the panic function is engaged and the system is shut down in emergency mode. This happens because the internal kernel structures have been changed and the new buffer size is insufficient to store a newly-generated ICMP packet. To cause this, the IP packet must satisfy the following criteria:

  • The size of the IP header should be 60 bytes.
  • The size of the  IP payload should be at least 65 bytes
  • There should be errors in the IP options (invalid size of option, class, etc.)

Example of packet that cause a crash

At first glance it is not obvious how this bug could be exploited effectively. However, a true professional can easily use it to break down a user's device or even interrupt the work of a corporate network. Usually this kind of incorrect packet would be dropped by routers or firewalls but we discovered several combinations of incorrect IP options that can pass through the Internet routers.

This vulnerability no longer exists in OS X 10.10.3 and iOS 8.3. In addition, users of Kaspersky Lab's products are secured against this vulnerability in OS X 10.10 by the Network Attack Blocker feature. Starting from Kaspersky Internet Security for Mac 15.0, this threat is detected as DoS.OSX.Yosemite.ICMP.Error.exploit.

The Banking Trojan Emotet: Detailed Analysis

Thu, 04/09/2015 - 10:00

Introduction

In the summer of 2014, the company Trend Micro announced the detection of a new threat - the banking Trojan Emotet.  The description indicated that the malware could steal bank account details by intercepting traffic.  We call this modification version 1.

In the autumn of that year a new version of Emotet was found.  It caught our attention for the following reasons:

  • The developers of this Trojan had begun to use technology that stole money automatically from victims' bank accounts - so called "Automatic Transfer System (ATS)".
  • The Trojan had a modular structure: it contained its own installation module, a banking module, a spam bot module, a module for stealing address books from MS Outlook and a module for organizing DDoS attacks (Nitol DDoS bot).
  • The creators made a significant effort to remain unnoticed: they didn't attack users in the RU zone but targeted the clients of a small number of German and Austrian banks (other well-known banking Trojans are less discerning in their choice of target),and the domain name of the ATS server changed frequently (once or several times a day).

We are going to refer to this modification as Emotet version 2. The bot contains and transfers the numbers one and seven to the command and control center (C&C), which suggests that the Trojan's authors considers this variant to be version 1.7.

Both versions of the Trojan attacked clients of German and Austrian banks.

#Trojan #Emotet targeted the clients of a small number of German, Austrian and Swiss banks

Tweet

We closely monitored Emotet version 2.  In December 2014 it ceased activity and the command servers stopped responding to infected computers.  We recorded the last command sent from the command centers on 10/12/2014, at 11:33:43 Moscow time.

However, the thoroughness with which the authors had approached the development of this Trojan and the high level of automation in its operation, left little doubt that this was not the end of the story.  And so it turned out - after a short break in January 2015, Emotet reappeared!  We are calling this modification version 3 (the bot contains and transfers the numbers one and 16 to the C&C, which we assume means that the authors consider this variant to be version 1.16).

In essence, Emotet version 3 is not that different to version 2 - the main differences are designed to make the Trojan less visible. Of the changes we noted, we would like to highlight the following:

  • The Trojan has a new built-in public RSA key and, although the communication protocols with the command center are identical for Emotet versions 2 and 3, if the old key is used the bot does not receive the correct answer from the command center.
  • The ATS scripts are partially cleaned of debugging information and comments.
  • New targets! Emotet is now also targeting clients of Swiss banks.
  • There has been a slight change in the technology used to inject code into the address space of explorer.exe.  Version 2 used a classic model for code injection: OpenProcess+WriteProcessMemeory+CreateRemoteThread. Version 3 uses only two stages of the previous model:OpenProcess+WriteProcessMemory;  and the injected code is initiated with the help of modified code of the ZwClose function in the address space of the explorer.exe process, which is also achieved using WriteProcessMemory.
  • Emotet version 3 resists investigation: if the Trojan detects that it has been started in a virtual machine it functions as usual but uses a different address list for the command centers.  However, all these addresses are false and are used only to mislead investigators.
  • The Trojan contains very few lines of text:  all lines that could warn investigators are encrypted using RC4 and are decrypted in allocated memory directly before use and deleted after use.

On the whole, we formed the impression that the main techniques used in version 3 of the banking Trojan were developed "in the field" using version 2 as a basis, and with the addition of improved stealth techniques.

Kaspersky Lab products detect all versions of this Trojan as Trojan-Banker.Win32.Emotet.  We also detect the following  modulesof Emotet:

  • Module for modifying HTTP(S) traffic - Trojan-Banker.Win32.Emotet.
  • Spam module - Trojan.Win32.Emospam.
  • Module for the collection of email addresses - Trojan.Win32.Emograbber.
  • Module for stealing email account data - Trojan-PSW.Win32.Emostealer.
  • Module designed for organising DDoS attacks — Trojan.Win32.ServStart.

We have seen the last module used with other malware and assume that it was added to Emotet by a cryptor.  It is quite possible that Emotet's authors are totally unaware of the presence of this module in their malware.  Whatever the case may be, the command centers of this module do not respond and the module has not been updated (its compilation date is 19 October 2014).

Infection

We currently know of only one method of distribution for the Emotet banking Trojan: distribution of spam mailings that include malicious attachments or links.

The attached files are usually ZIP archives containing the Emotet loader.  The files in the archives have long names, e.g. rechnung_november_2014_11_0029302375471_03_44_0039938289.exe.  This is done on purpose: a user opening the archive in a standard Windows panel might not see the extension .exe, as the end of the file name might not be displayed.  Sometimes there is no attachment and the text in the main body of the email contains a link to a malicious executable file or archive.

#Emotet banking #Trojan is distributed of spam mailings that include malicious attachments or links

Tweet

Examples of emails used to spread Emotet are given below.

Version 2 (link to malware):

Version 2 (attached archive):

Version 3 (link to malware):


The emails we found are almost identical to ones from well-known companies – for example Deutsche Telekom AG and DHL International GmbH.  Even the images contained in the messages are loaded from the official servers telekom.de and dhl.com, respectively.

When the email contains a link to malware, it downloads it from the addresses of compromised legitimate sites:

hxxp://*******/82nBRaLiv (for version 2)
or from the addresses
hxxp://*******/dhl_paket_de_DE and hxxp://*******/dhl_paket_de_DE (for version 3).

In Emotet version 3, when addresses are contacted with the form hxxp://*/dhl_paket_de_DE, the user receives a ZIP archive of the following form hxxp://*/dhl_paket_de_DE_26401756290104624513.zip.

The archive contains an exe-file with a long name (to hide the extension) and a PDF document icon.

Loading the Trojan

The Trojan file is packed by a cryptor, the main purpose of which is to avoid detection by anti-virus programs.  After being started and processed by the cryptor, control is passed to the main Emotet module - the loader.  This has to embed itself in the system, link with the command server, download additional modules and then run them.

Consolidation in the system is fairly standard — Emotet version 2 saves itself in %APPDATA%\Identities with a random name of eight characters (for example — wlyqvago.exe); adds itself to the autoloader (HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Run) and  then deletes its source file with the help of a launched bat-file that is created in %APPDATA% with the name "ms[7_random_numbers].bat.

Emotet version 3 saves itself in %APPDATA%\Microsoft\ with a name in the format msdb%x.exe" (for example – C:\Documents and Settings\Administrator\Application Data\Microsoft\msdbfe1b033.exe); adds itself to the autoloader (HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Run) and then deletes itself with the help of a launched bat-file (which is created in %APPDATA%\del%x.bat).

After consolidating itself in the system, Emotet obtains a list of the names of all processes running and calculates a hash from the name of every function, comparing the resulting value with the hardcoded  0xB316A779 (this hash corresponds to the process explorer.exe).  In this way, Emotet locates the process into which to inject itself.  Further, the Trojan unpacks its main code and injects it into the process explorer.exe.

Communication with the command center

The main module of the Trojan, the loader, communicates with the C&C using RC4 encryption.

The port used by the loader is hardcoded into it - 8080.

Command center addresses

The IP addresses of Emotet's command-and-control servers are hardcoded into the bot. There are several of these – one of the version 2 samples that we analyzed included 30 (note that 3 addresses on the list below belong to well-known legitimate resources):

hxxp://109.123.78.10
hxxp://66.54.51.172
hxxp://108.161.128.103
hxxp://195.210.29.237
hxxp://5.35.249.46
hxxp://5.159.57.195
hxxp://206.210.70.175
hxxp://88.80.187.139
hxxp://188.93.174.136
hxxp://130.133.3.7
hxxp://162.144.79.192
hxxp://79.110.90.207
hxxp://72.18.204.17
hxxp://212.129.13.110
hxxp://66.228.61.248
hxxp://193.171.152.53
hxxp://129.187.254.237
hxxp://178.248.200.118
hxxp://133.242.19.182
hxxp://195.154.243.237
hxxp://80.237.133.77
hxxp://158.255.238.163
hxxp://91.198.174.192
hxxp://46.105.236.18
hxxp://205.186.139.105
hxxp://72.10.49.117
hxxp://133.242.54.221
hxxp://198.1.66.98
hxxp://148.251.11.107
hxxp://213.208.154.110

In the sample of version 3 we investigated there were 19 command centers:

hxxp://192.163.245.236
hxxp://88.80.189.50
hxxp://185.46.55.88
hxxp://173.255.248.34
hxxp://104.219.55.50
hxxp://200.159.128.19
hxxp://198.23.78.98
hxxp://70.32.92.133
hxxp://192.163.253.154
hxxp://192.138.21.214
hxxp://106.187.103.213
hxxp://162.144.80.214
hxxp://128.199.214.100
hxxp://69.167.152.111
hxxp://46.214.107.142
hxxp://195.154.176.172
hxxp://106.186.17.24
hxxp://74.207.247.144
hxxp://209.250.6.60

Communication with the C&C when run in a virtual machine

Emotet version 3 contains another list of "command center" addresses, as given below:

hxxp://142.34.138.90
hxxp://74.217.254.29
hxxp://212.48.85.224
hxxp://167.216.129.13
hxxp://91.194.151.38
hxxp://162.42.207.58
hxxp://104.28.17.67
hxxp://8.247.6.134
hxxp://5.9.189.24
hxxp://78.129.213.41
hxxp://184.86.225.91
hxxp://107.189.160.196
hxxp://88.208.193.123
hxxp://50.56.135.44
hxxp://184.106.3.194
hxxp://185.31.17.144
hxxp://67.19.105.107
hxxp://218.185.224.231

The Trojan tries to contact these addresses if it detects that it is being run in a virtual machine.  But none of the addresses correspond to the bot's command centers, and the bot is therefore unsuccessful in trying to establish contact with them. This is probably done to confuse any investigators and give them the impression that the Trojan command centers are dead.  A similar approach was used previously in the high-profile banking Trojan, Citadel.

#Trojan #Emotet tries to contact the wrong addresses of the C&C if it is being run in a virtual machine

Tweet

The detection of a virtual machine is organized quite simply — by the names of processes that are usual for various virtual machines.  The following algorithm is used to calculate a hash value from the name of every process in the system:

Algorithm for calculation of a hash value from a process name

The resulting hash value is then compared with a list of values hardcoded into the Trojan:

Hashes from the names of processes used for the detection of virtual machines

We derived the names of the processes for several hashes. For example, hash 0xBCF398B5 corresponds to the process vboxservice.exe, hash 0x2C967737 to the process vmacthlp.exe, hash 0xE3EBFE44 to the process vmtoolsd.exe, and 0x61F15513 to the process vboxtray.exe.

Data transferred

A request to the command center appears in the traffic as follows (the example given is from version 2, but a version 3 request looks the same):

Dialogue between the Emotet bot and its command center

The URL-path that the bot communicates with appears as follows: /722ffc5e/355c7a0a/, where 722ffc5e is a number calculated on the basis of information from the access marker of the user, and  0x355c7a0a = 0x722ffc5e xor 0x47738654 (the value 0x47738654 is hardcoded into the bot).

The data sent by the bot and the command center are encrypted using RC4 and the answers received from the command center are signed with a digital signature.  Probably this is done to make it difficult to seize control over the botnet: in order for the bot to accept a packet it must be signed and for that it is necessary to know the secret key.

There is a public RSA key in the body of the bot. In PEM format for version 2 it appears as follows:

PEM representation of the open RSA key coded into the bot in version 2

As noted above, in version 3 the key changed.  In PEM format it looks like this:

PEM representation of the open RSA key coded into the bot in version 3

A packet sent to the server is made up as follows:

  • A request is generated containing the identifier of the infected computer, a value presumably indicating the version of the bot; information about the system (OS version, service pack version, product type); a hardcoded dword (value in the investigated sample — seven); control sums for the banker module; and information about the web-injects.  Information about the web-injects contains: a page address (with jokers), into which the injection is needed; data coming before the injected data; data coming after injected data; and injected data.
  • An SHA1 hash is calculated from the generated request.
  • The request is encrypted with a randomly generated 128 bit RC4 key.
  • The generated RC4 key is encrypted using the public RSA key.
  • The total packet is the concatenation of the results obtained at steps 4, 2 and 3.

The request packet can be represented by the following diagram:

Structure of a request from the bot to the server

In response the server sends a packet with the following structure:

Structure of the server's answer to the bot

The answer can contain information about the Emotet web-injects, Emotet modules and links for loading external modules (for example a spam bot or an updated loader).

Modules

Like most modern banking Trojans, Emotet has a modular structure.  To date we have detected the following modules:

Name Description Method of delivery to infected system loader loader In spam emails or by downloading via a link from a compromised site (for updates). nitol-like-ddos-module DDoS-bot mss Spam module Downloaded from compromised sites by the loader module. email_accounts_grabber Email account grabber, uses Mail PassView – a legitimate program designed for recovering forgotten passwords and mail accounts Received by the loader module in the answer packet from the command center. banker Module for modifying HTTP(S)-traffic Received by the loader module in the answer packet from the command center. outlook_grabber Outlook address book grabber Received by the loader module in the answer packet from the command center.

Several modules can work independently of the loader module, as they don't need to import anything from it.

The whole arrangement of the bot is evidence of a high level of automation: new email addresses are collected automatically from the victims' address books, spam with the Emotet loader is sent automatically, and money is transferred automatically from the user.  Operator participation is kept to a minimum.

As an example, here is the report of the outlook_grabber module sent to the attacker (from Emotet version 2) with a stolen Outlook address book:

A stolen Outlook address book, transferred to the criminals' server

One positive note is that when trying to contact one of the attackers' servers an answer is obtained containing "X-Sinkhole: Malware sinkhole", meaning that the stolen data will not reach the criminals — this domain, which is used by Emotet version 2, is no longer controlled by the authors of the Trojan.

However, for version 3 things are different.  This is how the report of the email_accounts_grabber module appears for Emotet version 3:

Report containing data about the user's email accounts

It is clear that the server answers "200 OK". This means that the criminals have successfully received the data.

Stand and Deliver!

Information about the data for injection into the page that is received by Emotet after unpacking appears as follows:

Decrypted data on the web-injects of Emotet version 2

Decrypted data in the web-injects of Emotet version 3

The significant difference in data on injects between the two versions is as follows: Emotet version 3 is aimed at the clients of Swiss credit organizations.  To date we have not seen scripts for the automatic stealing of money from clients' accounts in these credit organizations but we are certain that such scripts will be written soon.

Although individual fragments of HTML code in the decrypted packet can be read easily, understanding the rules for use of the web-injects from the deciphered data is difficult.  Below, in JSON format, several web-inject rules are given for one target — the site of a German bank (Emotet version 2).

The web-inject rules for the site of a German bank (Emotet version 2)

The use of this web-inject leads to the creation of a new element of type 'div', which will have the size of the whole visible page, and to the addition of a new script in the HTML document.  In the example given the script is loaded from the address hxxps://*******.eu/birten/luck.php?lnk=js&id=44.

And an analogous view of several inject rules for a new target — the site of a large Austrian bank (Emotet version 3).

The web-inject rules for the site of an Austrian bank (Emotet version 3)

It is clear that the configuration file with the web-injects has a classic structure, using fields conventionally called  data_before, data_after and data_inject.

It should be noted that the address of the host on which the file luck.php (for version 2) and a_00.php (for version 3) is located is changed frequently.  The rest of the address of the script is constant.

If the investigator tries the script directly, only an error message is received.  However, in a real attack when the line

is added to the real bank page, the script loads successfully.

This happens because the criminals' server checks the "Referer" field of the header of the HTTP request and sends the script only if the request came from a page of one of the banks attacked by Emotet.

Having supplied the necessary Referrer one can easily obtain the script code.

At Kaspersky Lab we obtained scripts designed for injection into the pages of the attacked banks.

Table 1.  Targets of Emotet version 2, types of attacks and the identification numbers of scripts loaded for carrying out these attacks.

Table 2. Targets of Emotet version 3, types of attacks and the identification numbers of scripts loaded for carrying out these attacks.

In one of the scripts of Emotet version 2 that was used to attack a German bank the comments contain the following line:

Artifact from the script for an attack on a German bank (Emotet version 2)

Clearly the script developers speak Russian.

Getting round two-factor authentication

The main purpose of the scripts looked at above is to carry out the illicit transfer of money from the user's account.  However the bot cannot independently get round the system of two-factor authentication (Chip TAN or SMS TAN), it needs the user's help.  To mislead the potential victim, social engineering techniques are used: the message injected into the webpage using the script informs the user that the site is introducing a new security system and normal operations cannot be continued until the user has tested it in the demo-regime.

False message about new security system

This is followed by a request to enter real data from the Chip TAN or SMS TAN to carry out a "test transfer":

And finally - congratulations that the task has been completed successfully:

In fact, instead of a test transfer the malicious script carries out a real transfer of money from the victim's account to the account of a nominated person — the so-called "drop", and the user themselves confirms this transfer using the Chip TAN or SMS TAN.

Details of the accounts for the transfer of the stolen money are not initially indicated in the script, but are received from the command server of the criminals using a special request.  In reply the command server returns a line with information about the "drop" for each specific transaction.  In the comments in one script we found the following line:

Clearly the criminals tested this script with a transfer of 1500.9 EUR to a test account.

In addition, this script contained the following information about the drop:

In the corresponding script in Emotet version 3, designed to attack the same bank, we also found information on the drop, but this time another one:

Let's compare the fields JSON __DropParam and the fields in the legitimate form from a demo-access to the online system of the attacked bank.

Online banking form for transfer of money within Germany or in the SEPA zone

Table 3. Relationship between the drop data and the fields in the form for transfer of money and explanations of these fields

Name of fields in the __DropParam JSON Name of corresponding field in the form Translation Field contents name Empfängername Name of recipient Real name of drop who will receive the stolen money ibanorkonto IBAN/Konto-Nr. International bank account number/ account number Account number, international or local, to which money will be transferred bicorblz BIC/BLZ BIC or BLZ code International bank identification code or identification code used by German and Austrian banks (Bankleitzahl) description Verwendungszweck Purpose Purpose of payment amount Betrag Amount Transferred amount

The JSON __DropParam fields correspond to the fields in the form.

In this way the bot receives all the necessary information about the drop from its server and draws up a transfer to it, and the misled user confirms the transfer using the Chip TAN or SMS TAN and waves goodbye to their money. 

Conclusion

The Emotet Trojan is a highly automated and developing, territorially-targeted bank threat. Its small size, the dispersal methods used and the modular architecture, all make Emotet a very effective weapon for the cyber-criminal.

The #Emotet #Trojan is a highly automated and developing, territorially-targeted bank threat

Tweet

However this banking Trojan doesn't incorporate conceptually new technology and so the use of a modern anti-virus program can provide an effective defense against the threat.

Furthermore, the Trojan cannot function effectively without the participation of the user — the Emotet creators have actively used social engineering techniques to achieve their criminal ends.

And so the alertness and technical awareness of the user, together with the use of a modern anti-virus program can provide reliable protection against not only Emotet but other` new banking threats working in a similar way.

Some MD5 hashes

Emotet version 2:
7c401bde8cafc5b745b9f65effbd588f
34c10ae0b87e3202fea252e25746c32d
9ab7b38da6eee714680adda3fdb08eb6
ae5fa7fa02e7a29e1b54f407b33108e7
1d4d5a1a66572955ad9e01bee0203c99
cdb4be5d62e049b6314058a8a27e975d
642a9becd99538738d6e0a7ebfbf2ef6
aca8bdbd8e79201892f8b46a3005744b
9b011c8f47d228d12160ca7cd6ca9c1f
6358fae78681a21dd26f63e8ac6148cc
ac49e85de3fced88e3e4ef78af173b37
c0f8b2e3f1989b93f749d8486ce6f609
1561359c46a2df408f9860b162e7e13b
a8ca1089d442543933456931240e6d45

Emotet version 3:
177ae9a7fc02130009762858ad182678
1a6fe1312339e26eb5f7444b89275ebf
257e82d6c0991d8bd2d6c8eee4c672c7
3855724146ff9cf8b9bbda26b828ff05
3bac5797afd28ac715605fa9e7306333
3d28b10bcf3999a1b317102109644bf1
4e2eb67aa36bd3da832e802cd5bdf8bc
4f81a713114c4180aeac8a6b082cee4d
52f05ee28bcfec95577d154c62d40100
772559c590cff62587c08a4a766744a7
806489b327e0f016fb1d509ae984f760
876a6a5252e0fc5c81cc852d5b167f2b
94fa5551d26c60a3ce9a10310c765a89
A5a86d5275fa2ccf8a55233959bc0274
b43afd499eb90cee778c22969f656cd2
b93a6ee991a9097dd8992efcacb3b2f7
ddd7cdbc60bd0cdf4c6d41329b43b4ce
e01954ac6d0009790c66b943e911063e
e49c549b95dbd8ebc0930ad3f147a4b9
ea804a986c02d734ad38ed0cb4d157a7

The author would like to express his thanks to Vladimir Kuskov, Oleg Kupreev and Yury Namestnikov for their assistance in the preparation of this article.

A flawed ransomware encryptor

Wed, 04/08/2015 - 06:00

In the middle of last year, my colleagues published a blogpost about a new generation of ransomware programs based on encryptor Trojans, and used the example of the Onion family (also known as CTB-Locker) to analyze how these programs work.

Last autumn, we discovered the first sample of an interesting new encryptor, TorLocker (this is the original name given by the creator); later on, TorLocker was used to launch an attack on Japanese users. When it was discovered on 24 October, 2014, the proactive components in Kaspersky Lab's products already detected this piece of malware; later on, it was assigned the verdict 'Trojan-Ransom.Win32.Scraper'.

Trojan-Ransom.Win32.Scraper encrypts the victim's documents and demands a ransom ($300 or greater) to decrypt them

Tweet

All the TorLocker samples that we have obtained belong to one of two versions: 1.0.1 (in English) or 2.0 (in English and Japanese.) There are only slight differences between them: 1) in the method employed to obfuscate code, and 2) in the sources used for additional modules: in the first version, the additional modules are extracted from the data section, while in the second version, they are downloaded from the Internet (from file hosting services or from compromised sites). Also in the second version, some strings were relocated from the data section into the code section, and dangling (redundant, not used) bytes emerged. The file encryption algorithm is the same in both versions.

Common features and peculiarities of this malware family

Our analysis has shown that Trojan-Ransom.Win32.Scraper was presumably written in assembler, which is unusual for this type of malware. The Trojan uses the Tor network to contact its "owners" – something that is apparently becoming a norm for the new generation of ransomware – and the proxy server polipo. This piece of malware often lands on users' computers via the Andromeda botnet.

Trojan-Ransom.Win32.Scraper encrypts the victim's documents and demands a ransom ($300 or greater) to decrypt them. If the malware gets deleted by a security product after the files are encrypted, the Trojan installs bright red wallpaper on the Desktop, containing a link to its executable file. Thus, users have a chance to re-install the Trojan and report to its owners that they have paid the ransom: to do so, users need to enter payment details in a dedicated TorLocker window. This data will be sent to the C&C server which will either reply with a private RSA key or notify that there was no payment.

This typical representative of the Scraper family is packed with UPX. The data section is additionally encrypted with AES with a 256-bit key. In the code section, between the assembler instructions, there are a large number of redundant bytes that are not used in any way.

The redundant bytes in the encryptor's body

The method of submitting string arguments to functions is just as unusual. The strings are located directly in the code section; in order to submit a string as an argument to a function, the pointer to that string is placed into the stack by way of calling (using the 'call' instruction) the instruction following the string. As a result, the return address (which is identical to the pointer to the string) is placed into the stack:

Handling string constants as arguments to functions

Operating principles

Once launched, the Trojan starts by decrypting its data section with a 256-bit AES key. The first 4 bytes of this key are used as a sample ID, added to the end of the encrypted files. Then the Trojan is copied to a temporary folder, and a registry key for that copy's autorun is created in the following registry section:

[HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Run]

Next the Trojan creates several threads to do the following:

  • Search for and terminate the taskmgr.exe, regedit.exe, procexp.exe, procexp64.exe processes.
  • Delete all system recovery points.
  • Encrypt the user's office documents, video and audio files, images, archives, databases, backup copies, virtual machines encryption keys, certificates and other files on all hard and network drives, except files located in the folders %windir%, %temp%. The names and extensions of encrypted files remain unchanged.
  • Here is the complete list of file extensions that are encrypted:
    .3gp .7z .accdb .ai .aiff .arw .avi .backup .bay .bin .blend .cdr .cer .cr2 .crt .crw .dat .dbf .dcr .der .dit .dng .doc .docm .docx .dwg .dxf .dxg .edb .eps .erf .flac .gif .hdd .indd .jpe .jpg .jpeg .kdc .kwm .log .m2ts .m4p .mdb .mdf .mef .mkv .mov .mp3 .mp4 .mpg .mpeg .mrw .ndf .nef .nrw .nvram .odb .odm .odp .ods .odt .ogg .orf .p12 .p7b .p7c .pdd .pdf .pef .pem .pfx .pif .png .ppt .pptm .pptx .psd .pst .ptx .pwm .qcow .qcow2 .qed .r3d .raf .rar .raw .rtf .rvt .rw2 .rwl .sav .sql .srf .srw .stm .txt .vbox .vdi .vhd .vhdx .vmdk .vmsd .vmx .vmxf .vob .wav .wb2 .wma .wmv .wpd .wps .xlk .xls .xlsb .xlsm .xlsx .zip
  • Extract a BMP image, save it to a temporary folder and then set it as desktop wallpaper:
  • Download tor.exe and polipo.exe, the files required to communicate with C&C servers, from the links specified in the Trojan's configuration (in the case of TorLocker 2.0) or extract them from the data section (in case of TorLocker 1.0). Then tor.exe is launched with the following arguments: tor.exe -SOCKSPort 9150 -AvoidDiskWrites 1 -ExcludeSingleHopRelays 0 -FascistFirewall 1 -DirReqStatistics 0

    polipo.exe is launched in the following configuration:

    127.0.0.1:57223 proxyPort = 57223 socksParentProxy = 127.0.0.1:9150 socksProxyType = socks5
  • Create a GUI window demanding that the victim pays the creators of Trojan-Ransom.Win32.Scraper and display the window in the top left corner of the screen. It supports payment via BitCoin, UKash and PaySafeCard.

    To encourage the user to pay the ransom to the Trojan's owners faster, the Trojan threatens to delete the private key required to decrypt the files if the user fails to send the money within a certain time period. In reality, the RSA keys are not deleted. They are associated with the malware sample rather than with a specific user, so the same RSA key is used for several users at the same time.

  • The IP address of the victim computer is determined using www.iplocation.net, www.seuip.com.br, whatismyipaddress.com, or checkip.dyndns.org.
  • Establish a connection to the C&C server in the onion domain via the proxy server polipo 127.0.0.1:57223. If the victim user has paid the ransom to the extorters, then, after contacting the C&C server and sending the information about the client (the selected RSA key, the number of encrypted files, the client's IP address and ID, the selected method of payment and the number of the bank card), the Trojan then receives the private RSA key with which to decrypt the files – in this case, a file decryption thread is created. Otherwise, a message is sent that the payment has not been effected yet. In each sample of Trojan-Ransom.Win32.Scraper?, a few dozens C&C domain names are hardcoded; they are not updated and may lead to the same C&C server.
Encryption

When launching, Trojan-Ransom.Win32.Scraper chooses one of the 128 public RSA keys hardcoded in it, depending on the victim computer's name and the serial number of the logical drive. The number (n) of the public RSA key is calculated as following:

n = (VolumeSerialNumber * strlen(ComputerName)) mod 128,
where strlen(ComputerName) is the length of the computer's name, and VolumeSerialNumber is the serial number of the logical drive on which Winsow is installed.

Each sample contains its own set of public keys.

The user's files are encrypted with AES-256 with a randomly generated one-time key; an individual encryption key is created for each file. Then, a 512-byte service section is added to the end of each file, which consists of 32 bytes of padding, 4 bytes of the Trojan's identifier, and 476 bytes of the employed AES key encrypted with RSA-2048.

If the file size is greater than 512 MB + 1 byte, then of the first 512 MB of the file get encrypted. The encrypted data is written on top of the original, non-encrypted data; no new file is created, and the old file is not deleted.

The Structure of an encrypted file

The Trojan does not need Internet access to encrypt the files.

Packing

In order to obstruct the analysis, some of the detected samples of Trojan-Ransom.Win32.Scraper were additionally packed with the KazyLoader and KazyRootkit protectors along with UPX.

KazyLoader is a two-stage protector of executable files, written in .NET Framework. The protected executable is encrypted with AES, and then placed into the protector's assets section as a color palette of a BMP image.

The image decryption module is encrypted by XORing with one byte, then divided into parts and also placed into the protector assets section in the form of strings LOADER0, LOADER1, … LOADER272.

The KazyRootkit protector is also written in .NET Framework and has a feature that can conceal processes in the Task Manager (taskmgr.exe) and conceal registry keys in the Registry Editor (regedit.exe) by deleting strings from ListView GUI elements with the help of WinAPI. Depending on its configuration, the protector may shut down without unpacking the file embedded in it, if it detects any of Sandboxie, Wireshark, WPE PRO or a code emulator.

Although Scraper (TorLocker) encrypts all files with AES-256 + RSA-2048, in 70%+ cases they can be decrypted

Tweet

The file to be protected is encrypted by XORing with a certain key, and then injected into the protector's process. A large array of random bytes is stored in the protector's overlay.

Partnership program

Trojan-Ransom.Win32.Scraper's builder (i.e. the program with which to create new samples of the Trojan with specified configuration) is distributed via a partnership program and sold for a few bitcoins. We found two posts about selling the builder for TorLocker 2.0 in the 'Evolution' (now taken down) underground online store:


The published screenshot of the builder suggests that the cybercriminal can change some of the encryptor's settings, as follows:

  • Allow or block the launch of Task Manager or Process Explorer after infection;
  • Allow or block the use of payment systems like BitCoin, PaySafeCard and Ukash to pay the ransom;
  • Allow or block the removal of Windows recovery points;
  • Modify the links from which to download tor.exe and polipo.exe; modify the names of these files after they are downloaded.

A screenshot of the builder's window

On the underground e-store's website, there are 11 reviews of the vendor of the Trojan-Ransom.Win32.Scraper builder, posted between 8 May 2014 and 17 January 2015.

By way of advertisement, news links are published about successful attacks performed using Trojan-Ransom.Win32.Scraper.

A brief description of TorLocker's operating principles and a comparison with CryptoLocker is also provided.

Decryption

At the decryption stage, when the ransom payment is received, Trojan-Ransom.Win32.Scraper contacts the cybercriminals' C&C servers via the Tor network and the polipo proxy server, to receive a private RSA key. With this key, the Trojan decrypts the AES key for each encrypted file, and then decrypts the files.

Although Trojan-Ransom.Win32.Scraper encrypts all files with AES-256 + RSA-2048, in 70%+ cases they can be decrypted because of the errors made during the implementation of cryptography algorithms. To restore the original files, Kaspersky Lab has developed the ScraperDecryptor utility, which can be downloaded from Kaspersky Lab's technical support website.

Blockchain Technology Abuse: Time to Think About Fixes

Tue, 04/07/2015 - 06:57

Kaspersky Lab and INTERPOL recently presented research on how blockchain-based cryptocurrencies could be abused through the pollution of public decentralized databases with arbitrary data.  During our presentation at the BlackHat Asia conference in Singapore, we demonstrated the proof-of-concept using the Bitcoin network, but it's important to understand that any cryptocurrency that relies on blockchain technology can be abused in this way.

Blockchain-based cryptocurrencies could be abused through the pollution of p2p databases with arbitrary data

Tweet

Some believe that security researchers, especially those from the anti-malware industry, generally only publish threat reports after the discovery of a threat in the wild.  However, this is not always true.  Our current research focuses on potential future threats that could be prevented before cryptocurrencies are fully adopted and standardized. While we generally support the idea of blockchain-based innovations, we think that, as part of the security community, it is our duty to help developers make such technologies fit-for-purpose and sustainable.

Blockchainware, short for blockchain-based software, stores some of its executable code in the decentralized databases of cryptocurrency transactions. It is based on the idea of establishing a connection to the P2P networks of cryptocurrency enthusiasts, fetching information from transaction records and running it as code. Depending on the payload fetched from the network, it can be either benign or malicious.

The proof-of-concept code we demonstrated was a benign piece of software

Tweet

To ensure the accurate interpretation of our research, we would like to point out that in the anti-malware industry, there is a clear definition of what constitutes malware, and there are extremely strict policies in place that forbid any attempts to create or distribute malware. The proof-of-concept code we demonstrated was a benign piece of software that opened the Notepad application after getting a confirmation from the user.

So, what exactly did we demonstrate at BlackHat Asia?   See for yourself at:  https://www.youtube.com/watch?v=FNsqXHbeMco

As we pointed out during our presentation, possible solutions can be introduced at different layers. From the perspective of a company developing endpoint security solutions, we don't believe it's too much trouble to blacklist applications that load unpredictable external payload from a P2P network.

We believe that the value of solution development lies in its neutrality and decentralized decision-making

Tweet

However, from the perspective of the cryptocurrency network, it's still an open question. We are not the experts in this field, and are therefore not best placed to propose effective solutions.  We also don't want to promote any specific solution as we believe that the value of solution development (as in the case of Bitcoin) lies in its neutrality and decentralized decision-making.

That's why we suggest this is a project for the cryptocurrency community.

We don't promote any specific solution. We suggest this is a project for the cryptocurrency community

Tweet

As a starting point for opening a discussion in the community, we suggest looking for an opportunity to implement a network consensus/negotiation algorithm that will sustain the clean state of the blockchain.

Don't Feel Left Out: Ransomware for IT Security Enthusiasts!

Tue, 04/07/2015 - 05:45

Macros are so hot right now

It's getting dark outside and our favorite mail client beeps with excitement for a new missive in our inbox, something interesting perhaps? A rapid glimpse at the contents of the message should indicate that a malicious campaign will play the starring role in what follows. An included attachment reveals itself as a malicious document with password-protected embedded macros. Moreover, a quick analysis of the file shows that it's dropping an executable payload to the system, which further piques our interest in this devious sample:

After opening the file, and only once the victim has been lured into enabling macros,  a seemingly innocuous Word document is shown.

File metadata betrays the developer's rush in crafting this file, using the Russian language letters "фыв" to fill the tags section:

"фыв" corresponds to the "asd" letter combination on Latin keyboards so often used as mindless filler.

Delving into the code

The second stage malicious script containing the instructions is downloaded from a public entry hosted on Pastebin in base64 encoding mode.

The full instruction set is 101 lines long and at the time of writing it counts with more than 5k reads. So this seems like a reliable indicator of the number of potential infections by this malware.

It is important to mention that upon discovery of the initial malicious document, Virustotal showed a null detection rate (however, the executable payload itself was detected by Kaspersky as Trojan-Ransom.Win32.Foreign.mdst)

The decoded script looks like this:

The decoded base64 payload downloaded from Pastebin fetches a file that includes several tokens to be used by the beckoning VBS script. Each token represents a section of the code that needs to be called in a specific order to achieve infection. The sections are named using a generic convention such as 'text20', 'text21', 'stext1', etc. Using the 'Tort' function implemented in the VBS script module, the instructions are deobfuscated and then outputted for execution.

The payload Trojan-Ransom.Win32.Foreign.mdst connects to an onion-based domain via the Tor2Web service

Tweet

In the case of the ' ' section, we can find a PowerShell script being called using the '-noexit' option, which according to Microsoft's Technet documentation is commonly used when running scripts via the command prompt (cmd.exe) so as to avoid exiting after execution. It's worth mentioning the second parameter, which sets the execution policy to bypass mode. Interestingly, by using a simple command line option this malicious creation is able to bypass the PowerShell execution policy configured in the system.

The file set for execution by PowerShell is also set by the original VBS script. A simple yet annoying obfuscation is in charge of getting the final string to be passed as a parameter.

As per the instructions above, the 'currentFile' variable will be replaced by the value of Chr(34) or a quotation mark, and the value of the variables PH2, FL2 and another static text value. Both PH2 and FL2 variables are set at the beginning of the execution of the script, FL2 being the random text used to name several files inside a temporary location set by PH2.

Even though the mechanism is not very complex, we can see that the malware writers took any measures available to slow down analysis and hide the real purpose of their code, even if by virtue of being a script it should be human readable.

We already reported the abusive Pastebin URL.

Payload

The payload is a binary PE file (self-extracting archive or SFX) named "file.exe". Upon execution, "file.exe" is copied to "C:\Windows\System32\WinSrv32.exe" and deleted from its original calling location. Persistence in the infected system is obtained via a registry key written in the following branch "HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Run".

This payload connects to an onion-based domain via the Tor2Web service.

The mention of a hostname refers to the front-facing side of the um6fsdil5ecma5kf.onion domain that serves as a C2 of the payload malware.

Detection names for malware 239d4f67692a5883574e3c496d88979c logmein_coupon.doc Trojan-Downloader.MsWord.Agent.hz 41d605b3981f330bd893b2dfd6e1d890 file.exe Trojan-Ransom.Win32.Foreign.mdst

Sinkholing Volatile Cedar DGA Infrastructure

Tue, 03/31/2015 - 16:35

There is currently some buzz about the Volatile Cedar APT activity in the middle east, a group that deploys not only custom built RATs, but usb propagation components, as reported by Check Point [pdf].

One interesting feature of the backdoors used by this group is their ability to first connect to a set of static updater command and control (c2) servers, which then redirect to other c2. When they cannot connect to their hardcoded static c2, they fall back to a DGA algorithm, and cycle through other domains to connect with.

Statistics:

This particular actor's true impact seemed interesting, so we sinkholed some of their dynamically generated command and control infrastructure. These victim statistics present a somewhat surprising profile. Almost all of these victims are geolocated in Lebanon.

Victims checking in to DGA c2

Clearly, the bulk of the victims we observe are all communicating from ip ranges maintained by ISPs in Lebanon. And most of the other checkins appear to be research related. Almost all of the backdoors communicating with sinkholed domains are the main "explosion" backdoor. But, some of the victim systems in Lebanon communicating with our sinkhole are running the very rare "micro" backdoor written up in the paper: "Micro is a rare Explosive version. It can best be described as a completely different version of the Trojan, with similarities to the rest of Explosive "family" (such as configuration and code base). We believe that Micro is actually an old ancestor of Explosive, from which all other versions were developed. As in other versions, this version is also dependent on a self-developed DLL named "wnhelp.dll." They check in to edortntexplore.info with the URI "/micro/data/index.php?micro=4" over port 443.

While Volatile Cedar certainly does not have a high level of technological prowess, it appears that they have been effective at spreading their malware, much like the Madi APT we reported on mid-2012. Because the group is not known for spearphishing, IT administrators should be aware of their own publicly exposed attack surface like web applications, ftp servers, ssh servers, etc, and ensure they are not vulnerable to SQLi, SSI attacks, and other server side offensive activity.

Kaspersky Verdicts and MD5s:

Trojan.Win32.Explosion.a
981234d969a4c5e6edea50df009efedd

Trojan.Win32.Explosion.b
7031426fb851e93965a72902842b7c2c

Trojan.Win32.Explosion.c
6f11a67803e1299a22c77c8e24072b82

Trojan.Win32.Explosion.d
eb7042ad32f41c0e577b5b504c7558ea

Trojan.Win32.Explosion.e
61b11b9e6baae4f764722a808119ed0c

Trojan.Win32.Explosion.f
c7ac6193245b76cc8cebc2835ee13532
184320a057e455555e3be22e67663722

Trojan.Win32.Explosion.g
5d437eb2a22ec8f37139788f2087d45d

Trojan.Win32.Explosion.i
7dbc46559efafe8ec8446b836129598c

Trojan.Win32.Explosion.j
c898aed0ab4173cc3ac7d4849d06e7fa

Trojan.Win32.Explosion.k
9a5a99def615966ea05e3067057d6b37

Trojan.Win32.Explosion.l
1dcac3178a1b85d5179ce75eace04d10

Trojan.Win32.Explosion.m
22872f40f5aad3354bbf641fe90f2fd6

Trojan.Win32.Explosion.n
2b9106e8df3aa98c3654a4e0733d83e7

Trojan.Win32.Explosion.o
08c988d6cebdd55f3b123f2d9d5507a6

Trojan.Win32.Explosion.p
1d4b0fc476b7d20f1ef590bcaa78dc5d

Trojan.Win32.Explosion.q
c9a4317f1002fefcc7a250c3d76d4b01

Trojan.Win32.Explosion.r
4f8b989bc424a39649805b5b93318295

Trojan.Win32.Explosion.s
3f35c97e9e87472030b84ae1bc932ffc

Trojan.Win32.Explosion.t
7cd87c4976f1b34a0b060a23faddbd19

Trojan.Win32.Explosion.u
ea53e618432ca0c823fafc06dc60b726

Trojan.Win32.Explosion.v
034e4c62965f8d5dd5d5a2ce34a53ba9

Trojan.Win32.Explosion.w
5ca3ac2949022e5c77335f7e228db1d8

Trojan.Win32.Explosion.x
ab3d0c748ced69557f78b7071879e50a

Trojan.Win32.Explosion.y
5b505d0286378efcca4df38ed4a26c90

Trojan.Win32.Explosion.z
e6f874b7629b11a2f5ed3cc2c123f8b6

Trojan.Win32.Explosion.aa
306d243745ba53d09353b3b722d471b8

Trojan.Win32.Explosion.ab
740c47c663f5205365ae9fb08adfb127

Trojan.Win32.Explosion.ac
c19e91a91a2fa55e869c42a70da9a506

Trojan.Win32.Explosion.ad
edaca6fb1896a120237b2ce13f6bc3e6

Trojan.Win32.Explosion.ae
d2074d6273f41c34e8ba370aa9af46ad

Trojan.Win32.Explosion.af
66e2adf710261e925db588b5fac98ad8
29eca6286a01c0b684f7d5f0bfe0c0e6
2783cee3aac144175fef308fc768ea63
f58f03121eed899290ed70f4d19af307

Trojan.Win32.Agent.adsct
826b772c81f41505f96fc18e666b1acd

Trojan-Dropper.Win32.Dycler.vhp
44b5a3af895f31e22f6bc4eb66bd3eb7

??
96b1221ba725f1aaeaaa63f63cf04092

IoT Research – Smartbands

Tue, 03/31/2015 - 07:00
Summary

Nowadays technology helps the development of hardware and software tools to record and analyze different aspects of our lives. This opens up new ways of staying aware of lifestyle and aiming to improve our health and fitness. One of the big trends in this sphere are fitness trackers such as smartbands, which, in the most popular current format, are bundles consisting of a hardware device we carry on our wrist and a mobile phone application to control the device and gain insights into the recorded data. We're entrusting these gadgets with very personal and sensitive data about ourselves and letting them into dive into our very inmost self. This poses big questions for us as a security company:

  • What kind of data is being collected
  • What are the risks and where are they?
  • What other parties might be interested in getting hold of this information, what's the potential result?
  • How can users help to protect their data?

Tracking devices and their corresponding mobile applications from three leading vendors were inspected in this report to shed some light on the current state of security and privacy of wearable fitness trackers.

What is it all about? The quantified self, smartbands and what people want to achieve

We regularly measure aspects of our daily lives because we feel we have to, because it is human nature to want to stay safe. We typically set our goals for certain points in time and regularly check how well or badly we're performing.

Things we often measure:

  • Business: financial goals, project plans, salary
  • Health: weight, height, eyesight, body mass index
  • Sports: heartbeat, distance covered and altitude gain while cycling or running, average speed

But a movement known as 'quantified self' wants more. It wants to go beyond and off the beaten tracks. This movement has been around for years and people are getting together all over the world to exchange information, discuss their experiences and form a culture of self-tracking. They are searching for a healthier, more fulfilling life by measuring things in their daily routines that have been overlooked by traditional measuring schemes.

The healthy living angle of this is attracting a lot of attention these days. Most people work in offices and they only get exercise as they commute, go shopping or walk to the coffee machine. More and more people work from home and use online store to get the things they need delivered to them, so there is far less need to actually leave the house. At the same time people are more aware of their bodies than – both in terms of health and an attractive appearance.

There are several ways of measuring how healthy, fit and active we are. Heart beat monitors help us control our exercise and get hard facts about our condition. Speedometers help cyclists measuring the distance covered, what altitude gain they achieved and their average speed. But all these tools are limited. They are taken off after exercising so other everyday activities like walking or working aren't recorded. If we use multiple devices the data remains isolated on each machine and is never correlated.

We entrust fitness trackers with our personal data and invite them into our innermost self

Tweet

This is where smartbands come into play. These devices are meant to be worn on our wrist all day and night to record our level of activity and also the time and quality of sleep. This generation of devices still records single snapshots, but the high frequency of recording sets makes it look like dynamic stream. It's a bit like the difference between photography, which gathers single shots, and filming, which uses a constant stream of shots to create a dynamic image. By acquiring and correlating constant streams of different health related data, we get additional benefits and information about our daily life, some of which we may not have been aware of. This paints a more complete picture of our lifestyle.

Human nature also seeks improvement. Collecting and visualizing our activities in daily life and their effects on our body helps motivate us to set new high scores. With most smartband offerings, users can try to beat their own targets as well as competing with a broader audience of family members, friends, colleagues from work and other individuals from online training groups. These are connected by the eco-system of the vendor's cloud network or by sharing information on social networks.

Smartbands – what they are and how they work

Basic smartbands are wristbands with mostly rubber surfaces to withstand shocks and moisture. The technological heart of the device is either firmly embedded into the body of the smartband or created in the form of a capsule, which can be placed into the band. The latter format allows the user to change the band if it gets damaged or worn out over time.

Bluetooth module: The main interface to upload collected data to a smartphone app and download new instructions, like vibration alarm at a defined time. Vibration motor: Just as on a smartphone, the motor lets the device vibrate to notify users of certain events like low battery or a pre-defined alarm. Motion sensor: Similar to the motion sensors in smartphones, the sensor monitors gyroscopic and accelerating movements. Vendor-defined algorithms then translate the movement into understandable units like steps. Battery: The battery of basic smartbands usually takes 35 – 70 mAH, a very low charge compared to smartphones, which take 2000 – 4000 mAH. Since there are far fewer components and they are usually more energy efficient, smartbands can keep running for one to two weeks, depending on how much data is being collected and how often power-consuming features are used. Power/sync button: Most smartbands can be operated with a single button to power on/off and sync or pair the device with the mobile phone. Power jack: To recharge the device's battery via a USB adapter. Display: Basic smartbands offer a small LED or dot matrix display to show battery charge or essential information like time or the step count. Features

Different smartband offerings have similar features. They are all based on measuring the activity levels, longevity and quality of sleep, information on calorie balance and additional goodies.

Main features:

  • step counter and approximate distance covered
  • calorie consumption
  • sleep recorder (duration and quality)
  • self-defined fitness plan and a comparison with actual activity

More features:

  • Nutrition intake and comparison with calories burnt from activity
  • Friend list with texting functionality and comparison of activities
  • Smart alarm for gentle recovery phase, based on measured stages of sleep
  • Stopwatch
  • Training diagrams
  • Third party extensions, if offered
A closer look at the whole system What data is collected by these devices?

The fitness trackers examined in this paper offer very similar feature sets and there is a consensus among the vendors about the data that is collected by the apps.

Required:

  • Name (or nickname)
  • Birth date (or just birth year)
  • Height
  • Weight
  • Gender
  • E-mail address
  • Password for account

Optional:

  • Country
  • Training plan
  • Weight goal
  • Training goals (steps per day, hours of sleep)
  • Nutrition plan
  • Photo
  • Mood
  • Friends using the same fitness tracker
  • . . .

The apps automatically show the correct localization, taken from the active settings on the mobile phone. Units for weight and height can be adjusted, enabling users to choose between imperial and metric systems, but is initially pre-set according to the mobile phone's localization setting.

Some fitness trackers allow users to control what they share on their friend list, but not with the cloud service.

Collecting and processing the information

The data acquisition and processing is done in a chain comprising the smartband itself, a smartphone (usually Android or iOS based) or computer (running on Windows or OSX), the corresponding application to process the data and the vendor's cloud service to provide deeper insights and store historical data. In order to synchronize the individual components the system uses Bluetooth and the Internet (via 3G/4G, Wi-Fi or wired connection).

The continuous synchronization between the tracking device and mobile phone requires a steady Bluetooth connection. This can have a considerable impact on the battery time of the phone. However the tracker is able to store data without synchronization for anywhere between two and 30 days, depending on the device and the amount of information recorded. Most vendors recommend keeping Bluetooth enabled at all times to ensure the best user experience.

Stage 1: Record data and short term storage

Stage 2: Process and correlate data, send instructions to control smartband

Stage 3: long time storage, web based interface for better viewing and deeper investigation

Smartbands are currently in a state of transition. The popularity of the product is prompting new varieties to come to market, and demand is growing for different formats. The type of smartband we know at present will be known as basic smartbands; future generations will offer additional innate processing power instead of merely collecting data. Some companies already have plans for products like combined smartwatches and activity sensors including heart beat monitors.

The daily traffic for cloud synchronization is around 1 -2 MB per day, depending on the model, level of activity and which features are used. Users without mobile Internet flat rates should consider performing this task via Wi-Fi only.

Possible vectors of compromise

In general, the more devices and data transmissions between them are needed in a system, the greater the possibility of compromising the chain. Most smartband environments use the above-mentioned scheme. Other types of fitness trackers cut out the smartband and record the data on the smartphone itself or don't offer a cloud service. For these types some attack vectors are not applicable.

Synchronization between tracking device and mobile phone

The smartband is meant to be worn day and night; however, their owners may well take them off from time to time. Therefore it could be left unattended for a while and anyone with a compatible device and the appropriate app – which is usually free of charge – could theoretically synch with the device and gain access to the data it records. That data could potentially be delivered to a rogue smartphone whenever it is range.

The information from smartbands and fitness trackers includes highly personal details about an individual. These could be used against the user for:

  • Blackmail
  • Naming and shaming on the Internet

Other than that, thieves might also be interested in the victim's training schedule since it could alert them to times when the flat or home is left empty.

The good news is that each of the smartbands we reviewed features some kind of integrated protection against this risk. The apps signed out from the phone and notified the owner that the smartband had been disconnected. The only information available to a rogue user was the data collected that day, or since the last synchronization. However, since only a small fraction of today's smartband offerings were tested, this attack vector might still apply to other devices.

The bad news is, the protection mechanisms can be susceptible to attacks, as my colleague Roman Unuchek proved in his blog post "How I hacked my smart bracelet". He was able to compromise the authentication process and thereby read the tracker's recorded data as well as executing code on it. According to his research, sometimes it is even possible to hijack the device without the owner even knowing.

Synchronization between the mobile phone and the server

The synchronization between the smartphone apps and the cloud servers is a neuralgic point, since the data stream comprises both the data gathered and the credentials to access the user's account. When smartbands hit the market some years back, some curious security researchers dipped into the traffic; a great uproar soon followed as it emerged that many vendors had no encryption whatsoever in this process, meaning all data was transmitted in clear text, perfectly readable for anyone who came upon it.

The synchronization between the smartphone apps and the cloud servers is a neuralgic point

Tweet

Fortunately, all the vendors of smartbands tested in this paper did their homework, since all of them incorporated a form encryption in their apps (TLS/SSL). This way, it is no longer simple to sniff traffic over Wi-Fi.

Compromising the mobile phone

Mobile malware has been a hot topic in recent years, with the number of new samples increasing in an almost exponential fashion. In the period from 2004-13 Kaspersky Lab analyzed almost 200,000 mobile malware code samples. In 2014 alone there was an additional stream of 295,539 samples. However, this doesn't give the whole picture. These code samples are re-used and re-packaged: in 2014 we saw 4,643,582 mobile malware installation packs (on top of the 10,000,000 installation packs that had been seen in the period 2004-13). The number of mobile malware attacks per month increased tenfold – from 69,000 per month in August 2013 to 644,000 in March 2014.

All the vendors of smartbands tested in this paper incorporated a form encryption in their apps (TLS/SSL)

Tweet

The typical modus operandi of cybercriminals is to use legitimate apps or app names as a vehicle to spread their malicious creations – mainly on third party app sites. One mobile malware sample is usually packaged under just one installer package, but sometimes even a hundred could be used to increase the leverage and therefore spread it among different user groups.  Malicious fake apps for smartbands, asking the user for the login credentials and thereby hijacking the account and all the information on it are entirely plausible. In combination with other data from the compromised phone, such as GPS coordinates of check-in features from social networking apps, this would pretty much by 'game over'.

However, the daily life of a smartphone poses a far higher risk. These devices are especially prone to getting lost. For example the London Underground reported more than 15,000 phones were lost in its trains in 2013 [1]. Without a lock screen in place, all information is visible to anyone who finds a smartphone, and that includes the information stored in fitness trackers. None of the smartband apps tested for this report offered the opportunity of locking the app with separate pin.

Compromising the cloud service

Aside from targeting single devices and users, attackers could also aim for the cloud service itself and seek access to records from all users.

Sometimes not even sophisticated hacking skills are needed, as one leading smartband vendor's user portal proved in 2011. All the user profiles were indexed by a popular search engine, making it easy to simply search the Internet for specific expressions that were only found in these profiles. Back then, users had the option to make their profile "private" but they were set to "public" by default. In addition, users could manually enter descriptions for their activities and certain timeframes, e.g. to find out what is most helpful when trying to lose weight. This meant that even the most private kind of "activity" was publically visible for everyone to see, together with information on longevity and how many calories were burned [2]. The vendor subsequently took action to prevent this. This case highlights how easily information and privacy leaks can result from misconfiguration and/or lax privacy policies.

Tracker's users have the option to make their profile "private", but they're set to public by default

Tweet

One smartband vendor's API allows users to access their data via a user ID and the serial number of the smartband, as well as the more traditional username-and-password combo. However, if a third party has the required information, the essential data can be downloaded without the user's knowledge.

In 2014 we saw numerous class A exploits like Shellshock or Heartbleed targeting web servers. These attacks were performed in a scripted fashion to IP addresses throughout the world by numerous gangs. It is still not clear how much data was gathered in these mega breaches, nor what the overall effect will be. Cloud services are not exempt from attacks like this and are seen as a lucrative target. It's only a matter of time until the next big exploit is found.

Other potential traps

According to research performed by the Massachusetts Institute of Technology, one smartband is notorious for scanning the user's environment for other Bluetooth-enabled devices, like computers, mobiles, other smartbands etc. As well as gathering the addresses of these devices, it also passes them to the vendor's servers via the smartband's phone application. This way, the vendor is potentially able to create a profile of each user's infrastructure environment.

In addition, the smartband itself uses BTLE (Bluetooth Low Energy), which makes it possible to change the device's address from time to time to avoid tracking the wearer. However, the vendor chose not to use this feature.

Fake smartband apps, asking for login credentials, are entirely plausible

Tweet

One tested smartband app invited the user to install additional apps from third parties to integrate and associate the collected data for deeper insights into the state of the users' health and activity. Possible extensions include correlating the standard data with GPS recording during workouts, dedicated apps for further visualizations, apps offering additional weight control related models, apps encouraging the user to eat more healthily (e.g. more fruit) and even offering financial incentives if all goals have been completed, paid by users who didn't meet their own targets.

If integrated, user automatically agrees to share this kind of data with the supplier.

The last potential trap has been a classic for decades. People tend to reuse their passwords over and over out of convenience. Most people have a main e-mail address, which also serves as a username on many websites and services. Now if one these accounts is compromised due to a server side breach (and we read of these breaches almost every week) or a malware infection stealing the login credentials on one of the machines, this means the other accounts using the same password are in massive danger. It is widely known that cybercriminals try these credentials on many of the big web portals like online shops, online payment systems, social networks and anything else that might turn a cash profit in the digital underground.

Commercial models around smartbands, fitness trackers and other gadgets

The flood of personal information, gathered by millions of users of smartbands and other wearables, whets the appetite of others as well as cybercriminals.

This kind of information is highly valuable to companies and institutions in different sectors.

Insurance companies

Insurance companies are based around risk estimation. To do this effectively, data has to be collected and evaluated to calculate the appropriate premiums from customers. The better the data, the better companies can manage their business. This is where fitness trackers come into play. What data could possibly be better than actual data streaming in real time from the customers themselves? At the time of writing some insurance companies are launching special programs for customers who are willing to share the information gathered from their fitness trackers. In return, financial incentives are offered to customers who prove to have a healthy lifestyle, as well as vouchers for travel and additional fitness courses [4].

None of the smartband apps tested offered the opportunity to lock the app with a PIN

Tweet

What could possibly go wrong? This scheme could potentially backfire. Imagine a keen fitness enthusiast who is also not averse to extreme sports. What if the tracking device and smartphone regularly transmit data about driving to an infamously dangerous mountain bike downhill track? GPS data sent from the smartphone and additional "step" count, coming from the rocks beneath the tire while riding at 40 kilometers per hour down the hill prove that someone didn't just go there to be a spectator, and this might displease the insurer. It could result in increased insurance costs based on the customer's allegedly higher risks. Depending on the legal situation in different regions in the world there's also a chance that insurance companies would refuse to insure high-risk clients because of the data recorded by their tracking devices.

Apart from fitness trackers, there are other gadgets and apps being developed to optimize the quantified self, like toothbrushes with integrated sensors to monitor the motion of the brush in three dimensions and a Bluetooth uplink to a dedicated smartphone application [5]. The app includes mini games to teach, motivate and reward people, especially children. It also tracks how often the teeth are being brushed and for how long. Again, insurers (dental insurance in particular) would be pleased to get their hands on this data.

Employers

Companies also discovered fitness trackers for their employees. There are already examples of employers offering these devices to their workforce to measure their health and motivate them towards a healthier lifestyle. British Petroleum (BP) introduced a "wellness program", in which employees are given points for reaching certain targets and incentives like health care premiums are offered [6]. Employees thinking about joining such program should thoroughly check the privacy policy and consider what potential consequences it might have.

Advertisement industry

There are almost no mobile apps on the market that offer users the option of disabling the flow of data into the cloud. As a result vendors quickly learn about your habits and your state of health. Depending on privacy policies, this enables them to tailor advertising based on the user's information and activity. Even within a general interest or activity, advertisements can focus on specific user groups: for example beginners could be offered running shoes and basic sportswear, whereas advanced athletes are shown advertisements for more expensive equipment, LED headlamps for night-time work outs or special sports nutrition. All offers can be adjusted for your local currency and targeted at the right gender and approximate size according to the weight and height set in the app.

Other parties

After the earthquake in Northern California the smartband vendor Jawbone published a diagram on their blog that showed the impact on sleep the event had in different areas around the epicenter [7]. All data was collected from thousands of customers, aggregated and presented in an anonymized form. The data enabled Jawbone to come up with a new format to show the actual impact of the earthquake on people rather than approximate seismographic ratings for surrounding areas. The graph appeared on many news sites around the globe.

Personal information gathered by millions of smartband users whets the appetite of cybercriminals

Tweet

The year 2014 marked the first time that data recordings from smartbands were used in court, opening the way for future cases. In this case the woman in question freely provided her data to prove that her injuries from a car accident limited her activities. Her information was compared with other women of her age using a third party [8]. In this case the use of data was not controversial – the woman provided her data freely to prove her point. It is important for smartband users to remember that vendors usually include a clause in their user agreements and privacy policies to make it clear that they can disclose information in response to a court order. It is also important to understand that the gathered data won't necessarily be kept in the country where it was recorded, but could also be used in foreign countries with a different jurisdiction.

According to researchers from the Hebrew University of Jerusalem it is possible identify individuals by the distinct shake of their GoPro cameras, worn on the head, from a sample of only a few seconds [9]. This raises the question whether algorithms along these lines could allow individual smartband users to be identified by their activity and sleep patterns.

Is there a more private way to keep track of your fitness?

More private alternatives (read: self-sufficient) to smartbands include pedometers and fitness tracker apps. Both options can act as single device systems and thereby cut off potential vectors of compromise that affect common smartband systems.

Fitness tracker apps commonly use an internal gyroscopic sensor and accelerometer to keep track of activities.  Tracker apps lack the sensors to measure people's sleep and also do without some other features of smart devices. Dedicated step counter devices, called pedometers, offer a similar feature set, but are easy on the smartphones' battery. Some offerings can be synced with a smartphone, others are completely self-sufficient. They can be carried it in a pocket or clipped onto your belt.

Advice for users of smartbands

To minimize the risks of your data being compromised, there are several pieces of advice to follow. Many of them apply not only to smartband users but to anyone using apps that store personal information:

  • Only use features you really need and avoid giving out any personal information that you would not want to store  in the cloud
  • Use a strong and unique password for each account
  • Lock the home screen on your smartphone and use access protection
  • Encrypt your phone if possible
  • Use security solutions for all devices, if available
  • Read the license agreements of applications and pay close attention to how personal information might be used by the service
  • Install app and operating system updates when available
  • Uninstall/Delete applications that are not needed anymore
  • Turn off the Bluetooth and location services on phones when not needed (this also preserves battery time)
Conclusion

Smartbands have been around for almost a decade by now, so they are almost senior citizens compared with many gadgets. While some old security issues like absence of encryption or public indexing of user profiles have been fixed, they show that security is still an afterthought for many companies. Security is also a process; vulnerabilities in drivers, protocols and the whole server ecosystem are found more and more frequently, vendors need to monitor vulnerabilities and the exploit landscape and quickly patch their software on both the client side (smartphone apps) and the server side (cloud service) to secure the customer data.

Security, though, depends on both makers and users alike. Everyone involved must understand the value and sensitivity of the user data collected by the fitness tracker. Normally when a breach involving personal data happens, data like names, mail addresses, birthdates, credit card information or passwords are affected. In this context, the information is even more personal. It contains health and body related data, including details that someone would normally confide only to a handful of very close people – or possibly even the doctor alone.

Smartband vendors are sitting on a goldmine of information that would be of great value to third parties in its anonymized form and even more attractive in a user-specific context. But if vendors decided to give out this data in either format (and risk losing their users' trust), third parties need to be cautious about the data. After all, what is to stop users attaching a smartband to a hyperactive pet dog and using that to get preferential 'active lifestyle' rates from an insurance company?

Although smartbands are relatively old technology, they are still part of the breeding ground for devices and services that trade on quantifying ourselves. New kinds of devices are coming up, integrating old technology and combining them with new innovation. Gadgets like smartwatches and Google's Glass are examples of how the future might shape up in this area.

Appendix: Resources

(1) More than 15,000 lost mobile phones on London Underground pose security risks
http://www.v3.co.uk/v3-uk/news/2318727/more-than-15-000-lost-mobile-phones-on-london-underground-pose-security-risks

(2) Dear Fitbit users, kudos on the 30 minute of vigorous sex activity last night
http://gizmodo.com/5817784/dear-fitbit-users-kudos-on-the-30-minutes-of-vigorous-sexual-activity-last-night

(3) Security Analysis of Wearable Fitness Devices (Fitbit)
https://courses.csail.mit.edu/6.857/2014/files/17-cyrbritt-webbhorn-specter-dmiao-hacking-fitbit.pdf

(4) Insurance company Generali wants to collect fitness data from customers (German)
http://www.heise.de/newsticker/meldung/Neue-Krankenversicherung-Generali-will-Fitnessdaten-von-Versicherten-sammeln-2461512.html

(5) Kolibree, Smart Tooth Brush
http://kolibree.com/en/

(6) Wearables at work mean big business, says Fitbit CEO
http://www.cnbc.com/id/101318809#

(7) How the Napa Earthquake Affected Bay Area Sleepers
https://jawbone.com/blog/napa-earthquake-effect-on-sleep/

(8) Fitbit Data now being used in the Court Room
http://www.forbes.com/sites/parmyolson/2014/11/16/fitbit-data-court-room-personal-injury-claim/

(9) Egocentric Video Biometrics
http://arxiv.org/abs/1411.7591

CanSecWest 2015: everything is hackable

Fri, 03/27/2015 - 09:48

Last week, we had the privilege to participate in and present at the 15th edition of CanSecWest in beautiful Vancouver, BC, along with its famous accompaniment, the ever famous Pwn2Own competition. Yes, once again all major browsers were hacked, but they were not alone! BIOS and UEFI, 4G modems, fingerprints, credentials, virtual machines, and operating systems were among the victim systems successfully hacked by our fellow presenters.

The event gathers a very technical audience with a shared interest in the most recent attacks and the presenters delivered with a variety of demos that showcased their intended vulnerabilities beautifully and thus reinforced the conclusion that digital voodoo can turn obscure and seemingly innocuous vulnerabilities into mind-numbingly cunning attacks.

One of the most discussed presentations, and certainly one of our favorites, showcased the power of BIOS and UEFI hacking: two guys, Corey Kallenberg and Xeno Kovah of Legbacore, armed with $2,000 and 4 weeks of hard work were able to show how a long list of vendor BIOSes were not only vulnerable but could successfully be loaded with LightEater, an SMM implant capable of pilfering sensitive information from Tails OS and even exfiltrating that information in such a way as to bypass the OS entirely. We clearly agree with their conclusion, it´s time to start taking a harder look at firmware!

Firmware insecurity: absence of evidence is not evidence of absence

One of the very possible attack is the well-known 'evil maid' or the 'border guard' approach: someone with physical access to your computer can just plug a small device (see below) and successfully reflash your system's BIOS, rewriting it with malicious code, without so much as booting up the system.

Press a button and in a few seconds the handy green light will indicate the BIOS is p0wned

Another very interesting presentation by Jan "starbug" Krissler showed how high resolution photos could bypass biometric authentication. Pictures acquired through high-resolution cameras from a safe distance amounted to the successful theft of fingerprints, faces, and irises used by current biometric systems for authentication. The distance can even be extended through the use of infrared imagery! We spent the talk imagining the breach possibilities as  an increasing number of ATMs  nowadays rely on biometric input.

Please authenticate access to your bank account using a password you can never change: your fingerprint

We also saw presentations on MacOS DLL (dylib) hijacking, userland exploits on iOS 8, attacks using Windows PowerShell, and even the installation of a bootkit in a 4G modem by simply sending an SMS! All sandwiched between explanations of the work of the ever fascinating Google Project Zero Team. In one of these, Chris Evans walked the audience through how a 'simple' crash caused by a call with a negative length became an exploit on Adobe Flash Player.

Our own presentation was a walkthrough of the misuse of whitelisted tools to further all kinds of attacks, from APTs and Targeted attacks to banking trojans and ransomware. This ongoing project is intended to highlight the faulty foundations of the whitelisting approach to security and how whitelisting alone simply won't protect you, from advanced and intermediary attackers alike! Stay tuned for a post on our findings.

In the end, we expanded our view as to the true breadth of vulnerable software and hardware. on which we depend daily. Security is a truly elusive state in an ecosystem composed of interwoven, dependent systems, each responding to the diverging priorities of a developer, an administrator, a user, and, of course, an attacker as well. The role of the security researcher that lives and breathes attack vectors and obscure vulnerabilities in search of the right digital voodoo has never been more important. And we can't help but echo the sentiments of Dragos Ruiu and our own Eugene Kaspersky in thanking CanSecWest for bringing all these researchers under one roof and one banner to share that digital voodoo and successfully stave off the balkanization of our industry just a while longer.

How I hacked my smart bracelet

Thu, 03/26/2015 - 07:00

This story began a few months ago when I got a popular brand of fitness bracelet. As this is a wearable device I installed Android Wear app, an application developed especially for wearable devices. This application easily connects to the fitness band.

However, there was something odd: the program could connect to a Nike+ Fuel Band SE, but my bracelet was another brand! It wasn't long before I realized my colleague had a Nike wristband – and he didn't even notice I had connected to his device.

After that I decided to do some research and find out how secure my wristband was.

Smart bracelets: communication with a smartphone

Today's market offers a lot of wristbands from other manufacturers. KSN provides the following statistics about the installation of Android-based applications to work with popular fitness trackers on mobile devices (the statistical data was obtained from KSN users who freely agreed to the transfer of this data).

The installation of Android-based applications designed to work with fitness trackers from different manufactures

Although this statistic demonstrates the popularity of Android applications (we cannot guarantee that the appropriate devices have users), to some extent it reflects the situation with the popularity of wearable devices.

To communicate with the smartphone most of these fitness bands use Bluetooth LE technology (also known as Bluetooth Smart). For us, this means that the devices connect in a different way from regular Bluetooth. There is no pairing password because most wristbands do not have a screen and/or a keyboard.

In some cases you can connect to a wearable device without the owner even knowing

Tweet

These wristbands use a GATT (Generic Attribute Profile) which means that every wearable device includes a set of services, each of which has a set of characteristics. Each characteristic contains a byte buffer and a list of descriptors, and each descriptor contains a value – a byte buffer.

In order to demonstrate this, I used some ready code from Android SDK, an example of an application that connects to Bluetooth LE devices. I did not have to write a single new line of code; I simply opened the existing project in Android Studio and pressed Start.

The screenshot above shows the result of my attempt to connect my fitness bracelet with the help of this application. Here we see the services and their characteristics. However, it is not easy to obtain data for my bracelet from the characteristics - it requires authentication in addition to the connection. In the case of some other devices I could read the data from the characteristics and their descriptors. This was probably the user data.

Scanning

So, using the example of the application from Android SDK I could connect to some devices. After that I have developed my own application which automatically searched for the Bluetooth LE devices attempting to connect to them and get their list of services.

Using this application I performed several scans.

  • Over two hours on the Moscow undeground subway system I could have connected to 19 devices: 11 FitBit and 8 Jawbone.
  • Over an hour in a gym in Bellevue, WA, USA I was able to connect to 25 devices: 20 Fitbit, and one each from Nike, Jawbone, Microsoft, Polar and Quans.
  • Over two hours at SAS2015 in Cancun, Mexico, I was able to connect to 10 fitness trackers: 3 Jawbone and 7 FitBit.

From just six hours of scanning I was able to connect to 54 devices despite two serious restrictions:

  1. Although the spec suggests the maximum distance for connections is 50 meters, in reality it's rarely possible to connect to a device more than 6m away.
  2. It seems that it is not possible to connect to a device that already has a connection to another phone. Thus if your wristband is connected to your phone, no one else can connect to it; it should not even be seen during scanning.

The second restriction should mean that when the wristband is connected to a smartphone, it cannot be attacked. This is not true though. And here is an example: while scanning with my app I was able to block the communication between my bracelet and its official application, even though they were connected.

It could be that the devices I found had never connected to a phone before or that the wristband was not connected to a smartphone while I was scanning (perhaps the Bluetooth on the phone was disabled). However it could also be that a pre-connected device was still available for connection despite the supposed restriction. Whatever the reason, potential fraudsters have ample opportunity to connect to fitness trackers.

However, in most cases, authentication is required in addition to the connection in order to gain access to the user data. Let's see how my bracelet's authentication process works.

My bracelet's authentication

To authenticate the bracelet on a smartphone the official application uses one of the four available services on the wristband. Each characteristic of each service is flagged with 'CharacteristicNotification' - this is how the app informs the wristband that it wants notifications of any change in this characteristic. Then the application gets a list of descriptors for each characteristic and sets the 'ENABLE_NOTIFICATION_VALUE' flag to inform the wristband that it wants notifications of any change in each descriptor.

After that one of the characteristics changes its value - the byte buffer. The application reads this buffer from the wristband: the 200f1f header and the byte array - let's call it authBytes.

The application creates a new array. Its first part is a constant array which is contained in the application and begins with 6dc351fd44; the second part of the new array is authBytes. The application receives the MD5 hash from the new array and sends it back to the device in the following structure:

  • Header (201210051f)
  • MD5
  • Verification byte

The application then sends to the device yet another array also found in this application.

After this wristband starts to vibrate and the user just needs to press the button to complete the authentication process.

With the official application the authentication process takes about 15 seconds. I have developed an application that requires only 4 seconds to make the wristband vibrate.

It is not difficult to make the user press a single button on the wristband. You just need to be persistent. You can keep trying authentication process over and over until the user finally presses the button or moves out of range.

From just six hours of scanning I was able to connect to 54 devices despite two serious restrictions

Tweet

After authentication is completed, the data on my bracelet can be accessed. Right now, wearable fitness devices do not contain much information. Typically, they have the number of steps, the phases of sleep, the pulse for the last hour or so. Approximately once an hour the app transfers this information from the wristband to the cloud.

After the authentication, it is easy to execute commands on the device. For example, to change the time you should send to the device the byte array beginning with f0020c and then the date in the form YYYY MM DD DW HH MM SS MSMSMSMS.

Things are even easier with the other fitness trackers: for some of them, part of the data is available immediately after the connection, while the application code for Nike is not even obfuscated and can be easy read (the results of one study can be found here).

Conclusion

The results of my research show that in some cases you can connect to a wearable device without the owner even knowing.

By hacking the bracelet I have the fraudster cannot get access to all user data as this is not stored on the wristband or in the phone - the official application regularly transfers information from the wristband to the cloud.

Fitness trackers are becoming more popular and offer a wider range of functions. Perhaps in the near future they will contain more sensors and hence much more user information, often medical data. However the creators of these devices seem to think very little about their safety.

Just imagine - if a wristband with the pulse sensor is hacked, store owners could look at your pulse rate while you are looking at the prices in the store. It might also become possible to find out how people react to advertising. Moreover, a hacked wearable with pulse sensor could be used as a lie detector.

The fraudster could take control of your wristband, make it vibrate constantly and demand money to make it stop

Tweet

Of course, there are more harmful actions that are more likely. For example, by using a Trojan-Ransom the fraudster could take control of your wristband, make it vibrate constantly and demand money to make it stop.

We reported our findings to my bracelet's vendor. The company's response defines the findings as a UX Bug and not a security issue. For ethical and security reasons we are not disclosing the name and the model of the bracelet this time. If you're worried about the possible consequences of cybercriminals exploiting the security issues we discovered, don't hesitate to contact the vendor of your fitness bracelet and ask if your product is affected by the method described in the article.

We also hope that this article will be helpful not only for users but also for vendors of the bracelets to make these devices safer from the IT Security perspective.

Analog OPSEC 101 – operational security in the physical world

Wed, 03/18/2015 - 06:00

For a long time we´ve been interested in operational security (OPSEC), and although you can find tons of cool technical tips about protecting digital information, we always felt that something was missing. After all, we live in a physical, or  analog world as well as a digital one, and we have encounters with other real people. After asking around, we found that one of the biggest worries of our technical community was how to behave during these interactions. So we decided to work on creating some realistic and easy to remember tips for exactly these situations.

Threat modeling

OPSEC is all about hiding information from your adversaries. We categorized our adversaries into just two groups: those who have resources and those who don´t. Plain and simple.

The first group comprises intelligence agencies, military organizations and the big bad boys. The second contains the rest. Important: no resources is not the same as no danger, but they are less able to track you unless you give away information for free.

Our tips are focused on encounters with the first group, since that is more likely to happen.

Recruitment

Agencies are always on the look-out for new assets to recruit – this is what they've been doing for centuries.

It all starts with the spotting process, identifying an asset who could meet their requirements based on the position and access to information. Next they profile the target, partly using OSINT. After that it's time to choose between the carrot and the stick, and pick out the most effective motivators on offer: money, blackmail, ideology, sex, etc.

Then some guy will approach us, maybe in person, maybe through LinkedIn. He'll probably pose as some businessman who will pay us a lot for nothing much, just a few easy reports from time to time.

When this happens we want to get to the Termination phase ASAP, ideally after being written off as a waste of time and effort.

We can just say "No", but they may keep increasing the pressure. On the other hand, we can refuse while providing alternatives, redirecting the request to another person ready to handle this.

Create a protocol for yourself and your organization in order to handle these situations effectively, minimizing the researcher ´s exposure. Be prepared in advance for situations where we are more vulnerable.

Borders

Crossing an international border can be one of the most vulnerable places. Somehow they are like a parallel dimension: although you are physically in one territory, the laws are just different, or maybe even non-existent.

We´ve learnt a few things regarding borders: there is always some exception to the law that officers might use in extreme scenarios. You can find legal advice here https://www.eff.org/wp/defending-privacy-us-border-guide-travelers-carrying-digital-devices>. However this is what you should NOT do:

  • Regardless of whether you consent to a search or not, do NOT stop the officer if he starts checking your stuff. This is a felony.
  • You don´t have to answer questions, but if you decide to do so, do not lie to the officer. Again, a felony.

This is our advice about how to react in a situation like this. These rules will provide you with peace of mind, help you stay calm and not freak out. Hopefully they will stop you overreacting, making things worse and talking too much, starting with: "I have nothing to hide, let me explain …".

  • Be cooperative.
  • Don´t make things worse.
  • Have your story prepared and be ready to back it up.
  • Golden rule: Don´t bring any valuable content with you! You should encrypt, upload and retrieve on arrival at your destination.
Other situations

Sometimes we could find ourselves going to a meeting in a strange country with a suspicion that something is not quite right. Some advice for this:

  • Don´t go alone.
  • Don´t rely on your host for transport
  • Plan exit routes and "safe" places, have your contacts ready.

In some cases the meeting itself won´t be the "trap"; it's just an excuse to get you to leave your computer in a known location the hotel, or in a cloakroom.

It is always a good idea to let someone know where you are going and tell them to react if you don´t ping them in a reasonable period of time. This also lets your adversaries know that you are ready – a simple casual comment will do the job.

Another concern is physical surveillance. To be honest, if this is done by sophisticated professionals there isn't much we can do about it and we probably won't even notice. But remember – don't try anything stupid; you're not James Bond. Acting like it's a movie can only make things worse.

If you are very concerned, escalate the situation and involve the person in your company who is responsible for dealing with local contacts. If you feel uncomfortable, move to a public place or directly move to your embassy.

Conclusions

You've probably already spotted a common theme in most of all these situations. First, keep calm and do not make things worse. You can rely on a third party to send in the cavalry when you need it. This is why your company should provide you with a single person to contact when you're in trouble. Also you might need international legal support.

However the key lesson is: do your homework. If you travel abroad, spend some time finding local contacts, get the telephone number and directions for your embassy, plan your meetings, let other people know where you are and make sure they are ready to act quickly in certain situations. Have your travel laptop ready and consider what information you bring with you. If you remember your lessons, you will be fine.

Yeti still Crouching in the Forest

Tue, 03/17/2015 - 03:53

Last July, we published details on Crouching Yeti (aka Energetic Bear), an advanced threat actor involved in several APT campaigns.

A quick summary:

  • Campaign status: Active
  • Discovery: January 2014
  • Targeted platforms: Windows
  • First known sample: 2010
  • Number of targets: 2,001-3,000
  • Top target countries : United States, Spain, Japan, Germany, France, Italy, Turkey, Ireland, Poland, China
  • Propagation method: Social engineering, Exploit, Watering hole attack, Trojanized software installers
  • Purpose/functions: Data theft
  • Special features : Interest in OPC/SCADA. Trojanized software used to administer remote OPC servers as well as modules to scan networks for OPC servers.
  • Targets: Industrial/machinery, Manufacturing, Pharmaceutical, Construction, Education, Information technology
  • Artifacts/attribution : Russian-speaking authors

This post is an update about the operational status of the campaign described in the original "Crouching Yeti" report.

Since the beginning of the research, we've been monitoring some of the C2 servers used by the components used in the attack – the Havex Trojan, the Sysmain Trojan and the ClientX backdoor. The following analysis is based on data gathered until March 04, 2015

C2 and victims:

Overall, we successfully monitored 69 C2 server (unique domains), receiving hits from 3699 victims (unique IDs of the Trojan/backdoor) connecting from 57796 different IP addresses. We gathered four additional C2s since the publication of the first report (65 in the last report).

Based on the graph below, the top five C2 servers share most of the unique victims :

Victims per C2

 

Although the trendline shows a decreasing number of hits on the C2, there are still >1.000 unique victim connections per day. These top five C2s with most of the victims coincides with the activity analyzed in the previous research and publication.

Another interesting figure is the number of hits by date which shows a decreasing trend:

The following figure shows the entire picture regarding Crouching Yeti victim country distribution including all the malware (Havex, ClientX,Sysmain) reporting to the C2s on which we have visibility. The graph contains the total dataset (inluding data for the previous report as well as the gathered during this period) and contains all the unique IP addresses observed. Be aware that there are some unique IDs using several IP addresses probably pertaining to infected computers used by travellers.

This shows the big (and updated) picture regarding Crouching Yeti victims by country. Spain, Poland and Greece are in the Top 3. Japan and especially the United States have significantly reduced position (less victims) since the last report, contrary to Poland and Italy that increased position remarkably (more victims reporting to the C2).

An additional representation of victim country distribution including the full dataset (all countries) :

Malware:

The most widely used Trojan on these C2 server is Havex with 3375 unqiue victims. Sysmain counts 314 and ClientX 10 (as in the last year's report). For Havex, version 024 is still the most widespread, followed by version 043. This is consistent with the trend observed in our last publication.

The following two graphs show the distribution of victims per malware type. We decided to divide the identified versions in two groups for purposes of clarity. The series names Report contains the data published in the first Crouching Yeti release (blue) and the Update (red) series contains the data analyzed.

During this period, the first subset shows an increase for almost all the included versions except for Havex-038 and Havex-01D which showed bigger activity in the first Crouching Yeti release . On the other hand, Havex-043 has the most significant increase during this period.

For the second subset, the picture looks pretty similar (global increase) except for Havex-01d which shows a decrease during this period.

Already before and also after the announcements around this actor other researcher digged into. Therefore the datasets are cleaned but may still include few research based non-victim systems.

The following graphs shows the operating system distribution amongst Havex victims during this period:

Apart from the increase of the category "unknown", there are no substantial differences when comparing the data analyzed in the first report :

In order to complement the data from the C2, we extracted some stats for the most relevant Trojans used by the Crouching Yeti operators. Almost all of them shows a residual impact during 2015. Nevertheless, we notice some very specific peaks during this month, especially for the Trojan.Win32.Ddex verdict. This component is a simple downloader with the functionality similar to the Havex component. All the detections are located within the Russian Federation.

In conclusion, the data analyzed during this period show us that Crouching Yeti's impact continues to increase in terms of infected victims reporting to the C2s, although internal data from KSN shows a different picture (residual number of infections). In this update, we did not see relevant changes in the infrastructure or in the C2 activity.

Taking into account the nature of this threat actor and the operational status of the infrastructure, it is likely the operators already switched infrastructure, techniques and targets.

We will continue to track this threat actor and providing updates accordingly.

Kaspersky Security Bulletin. Spam in 2014

Thu, 03/12/2015 - 08:00
The year in figures

According to Kaspersky Lab, in 2014

  • The proportion of spam in email flows was 66.76%, which is 2.84 percentage points lower than in 2013
  • 74.5% of spam emails were no more than 1 KB in size
  • 16.71% of spam was sent from the USA
  • Users in the USA were targeted by 9.8% of malicious emails, the largest share of any country
  • 260, 403,422 instances that triggered the "Antiphishing" system were recorded.
  • Brazil had the highest proportion of people attacked by phishers – 27.47% of all Kaspersky Lab users in the country faced at least one attack
  • Russia suffered the highest number of total phishing attacks, with 17.28% of the global total
  • 42.59% of phishing attacks targeted global portals that integrate many services accessed from a single account.
Popularity of mobile devices and spam

The popularity of mobile devices continues to grow, and this is affecting spam in email traffic: the number of advertising services that will spread spam on mobile devices is increasing, as are the number of offers addressed to the spammers who profit from these mailings. The popularity of mobile devices also makes them a valid vector for cyber-attack: email traffic now includes malicious imitations of emails sent from smartphones as well as fake notifications from popular mobile applications.

Adverts from/for "mobile" spammers

In 2014, spammers intensified their offers to distribute ads via SMS and popular IM services (WhatsApp, Viber, etc.). They use traditional spam mailings to help search for new customers, and the number of these adverts is also increasing.

Current email traffic also includes adverts addressed to "mobile" spammers: they are offered ready-made databases of phone numbers and other contact information that is designed to attract a specific target audience. These databases, in turn, are often generated with the help of mass mailings: the spammers send phishing emails which they use to collect personal data from victims.

Imitations of emails sent from mobile devices

Spam mailings simulating emails sent from mobile devices have become very popular. We came across such emails written in several languages; they mentioned iPad, iPhone, Samsung Galaxy and other models. These messages had one thing in common - short (sometimes non-existent) text and a signature reading "Sent from my iPhone". Typically, they contain links to phishing sites or malicious attachments.

Apparently, spammers think that an email with the attached file and a signature allegedly sent from the iPhone looks reliable. Indeed, the emails sent from mobile devices rarely use a complex template. And the senders often prefer to attach a file or insert a link rather than write a long text on the smartphone.

In some cases, the emails included an archive named to suggest it contained a photo. In fact, this was yet another way to distribute malware.

The emails sent allegedly from mobile devices often contained advertising links - most often to the sites illegally selling medications. Below is the example of one of these emails where the spammers used a few key words as the text of the message.

In order to bypass filtering, spammers often try to forge technical headers of the emails (Data, X-Mailer, Message-ID) to make them look like they were sent from mobile devices. However, when checked, the content of these headers is defined as incorrect.

Fake notifications from mobile applications

The widespread use of mobile devices has given rise to yet another phenomenon - spam that imitates notifications from different mobile applications such as WhatsApp and Viber. Users are accustomed to the synchronization of cross-platform applications, to the synchronization of contact data between applications and to different notifications from them, so many mobile device owners don't think twice about an email saying a message has allegedly arrived on their mobile messenger. This is a mistake: these mobile applications are not related to the user's email account in any way, which means that these emails are obviously fake.

For example, there can't be emails telling users that they've received an image via WhatsApp because registration on WhatsApp does not require an email address.

Moreover, the "picture" is packaged in an archive, which should also arouse suspicion: packaging an image does not offer any advantage while archives are often used to hide malicious attachments. And this is what we see in this case: the archive contains a malicious program.

Yet another example: a notification about a voice message allegedly sent via Hangouts contains a hyperlink disguised as the "Play" button. After clicking "Play", instead of hearing the voice message the user is sent to a compromised legitimate site from which integrated JavaScript redirects him to an advertising page.

The notification of the voice message supposedly sent via Viber contains a "Listen to Voice Message" button that initiates the download of a malicious archive.

World events in spam

2014 was rich in global events: the crisis in Ukraine, the Ebola epidemic, the Olympics in Sochi and the FIFA World Cup in Brazil. Each of them was used by the spammers to draw attention to their mass mailings.

The Olympic games and the World Cup

The Sochi Olympics and the FIFA World Cup in Brazil were the only sporting events that featured in spam flow. In both cases the majority of spam emails were written in the language of the country where the event took place, suggesting that the fraudsters' main targets were the locals.

The spam sent out shortly before these events contained a lot of mass mailing advertising products with the symbols of the tournament. In the case of the Olympic Games, in addition to Sochi merchandise the adverts offered products harking back to the 1980 Moscow Olympics.

"Nigerian" scammers also got involved. Before the Olympics they sent emails on behalf of the fans who asked for assistance in renting accommodation in Sochi or paying for various services. The "sports fans" were allegedly ready to transfer 850,000 euro to a person who would help them. The promise of a big reward was designed to encourage victims to overlook a few preliminary costs, but once the requested money was transferred the fraudsters disappeared without delivering the promised cash and rewards.

We also saw many fraudulent emails informing recipients that they had won an official FIFA World Cup lottery. Of course, to get his money, the 'winner' faced a few minor preliminary expenses. But, of course, these so-called winners would never see any money from a competition that they had never entered in the first place.

Similar emails are constantly sent out in the run-up to the big football championships.

Besides real adverts and fraudulent messages the pre-World Cup spam also contained malicious emails with links that allegedly led to websites where fans could buy tickets to the games.

Nelson Mandela's demise

Obviously, "Nigerian" scammers do not grieve for the demise of political leaders; for them these events are the ideal pretext to spin stories of a multi-million dollar will. Nelson Mandela's death in late 2013 unleashed a wave of "Nigerian" spam. The attackers introduced themselves as representatives of different funds and informed the recipient that he had been awarded a Mandela prize; the "bankers" offered to secretly divide the Mandela family's money, etc. In some cases, the emails contained the links to real news releases, hoping that this would make the message look more reliable.

Political events in Ukraine

Unstable political situations and military conflict is yet another source of inspiration for "Nigerian" spammers. We regularly come across mass mailings exploiting conflicts in different countries, mostly in the Middle East, but in 2014 the "Nigerian" scammers focused on Ukraine. The authors of fraudulent emails posed as the disgraced Ukrainian politicians and entrepreneurs looking for a way to smuggle their millions out of the country. There were also mass mailings written on behalf of Russian businessmen who had suffered due to sanctions.

Traditionally, "Nigerian" letters offered the recipients huge money for their help. Meanwhile, if victims entered into correspondence, the scammers conned money from them to cover different alleged expenses - duties, taxes, air tickets, hotel rooms and so on.

The Ebola virus

The Ebola epidemic also attracted the attention of spammers. "Nigerians" sent out emails on behalf of infected Africans who allegedly wanted to leave their fortune to charity. The fraudsters came up with a new twist on the story which invited recipients to participate as a guest at a World Health Organization conference. The proposed fee was 350,000 and a car for the job as a WHO representative in the UK.

Malware distributors exploited people's fear of this deadly disease and sent out emails on behalf of the WHO which contained a link to information on the measures to prevent Ebola infection. Later, the emails with the similar content appeared but this time the "information from the WHO" was packed in an attached archive.

In reality both the link and the attached archive contained a malicious program designed to steal the victim's data. In the above email it was Backdoor.Win32.DarkKomet.dtzn.

Spammer tricks

The techniques that have been actively used by the spammers in recent years can be called "classic".

One well-known spammer trick is the stock spam advertising shares of small companies. These emails are part of a stock fraud scheme, the so-called pump and dump spam. The idea is simple: the fraudsters buy cheap stock then send out email mailings advertising the chance to buy stock in a certain company at super low prices, taking advantage of the sharp rise in value expected in the near future. As a result, demand for the stock in the company rises, the prices are artificially inflated and the scammers sell off their stock in the company at a tidy profit. This fraud peaked in 2006-2007 but stock spam is still in use today.

According to Kaspersky Lab, 74.5% of #spam emails sent in 2014 were smaller than 1 KB in size #KLReport

Tweet

In 2013, stock spam only contained a brief text showing the current and expected share price of the company. Some mass mailings included an auto signature which promised that an anti-virus scan had been deployed. Moreover, the language of the signature matched the language of the geographical domain which hosted the recipient's e-mail (this common technique seeks to persuade the recipient of the legitimacy and security of the email). To enhance the chances of bypassing spam filters, the name of the company in a mass mailing was usually "noised" with the "_" symbol or gaps, and the text fragments varied.

In 2014, the design of fraudulent mass mailings advertising company shares changed – the spammers made the messages look more reliable. To bypass spam filters, they used some well-known tricks:

  1. Graphic spam. The advertising text is located within the picture and company logos are used. In a single mass mailing the content, color, font size or background color of the picture can vary. (Note that modern spam filters have long used graphic analyzers that can easily detect graphic spam)
  2. Junk text is inserted at the end of each message and is designed in different colors which do not always match the background. Fragments from literary works and quotes from Wikipedia are used. It is assumed that this method will make each email unique and cause spam filters to detect them as fragments of a literary work rather than a spam message.

Apparently, spammers are trying to compensate for their archaic methods with large volumes – hundreds of millions of these fraudulent emails are sent out.

However, spammers often use more advanced techniques to create "background noise" in the text. For example, they can "noise" the main text of the message even without affecting the readability of the message. To do this, HTML tags are used. The opening and closing tags are inserted into the main text of the message in HTML code. As a result, the user sees no changes in the message but the spam filter detects each email as a unique.

Statistics The proportion of spam in email traffic

In 2014, the proportion of spam in email traffic was 66.76%, which is 2.84 percentage points lower than in the previous year. Spam levels have fallen consistently from a peak of 85.2% in 2009. This is due to the fact that adverts for legal goods and services are abandoning spam in favor of more effective legal advertising platforms.

The proportion of spam in email traffic, 2014

In 2013 the share of spam in email traffic showed almost no variation from month to month but in 2014 there were some noticeable fluctuations, especially in the first half of the year. The lowest value of the year (63.5%) was registered in March. However this was immediately followed by the highest monthly figure of 71.1% in April. The second half of the year was more stable.

Sources of spam by country

Sources of spam by country, 2014

In 2013 China was the undisputed leader among the spam source countries. However in 2014, the percentage of unwanted mail originated from this country dropped by 17.44 pp. As a result, China fell to 3rd in the annual rating, overtaken by the USA (-1.08 pp) and Russia (+1.98 pp).

The Top 10 sources of spam include three western European countries: Germany (+2.79 pp), Spain (+2.56 pp) and France (+2.33 pp). Two Asian countries - South Korea (-10.45 pp) and Taiwan (-3.59 pp) – which occupied 3rd and 4th positions in 2013 moved down to 13th and 14th places respectively in 2014.

The size of spam emails

The size of spam emails in 2014

The number of super-short spam emails is growing: in 2014 77.26% of spam emails weighed in at under 1 KB, 2.76pp more than in 2013.

These emails usually contain links to advertising websites. To generate the text of the email the spammers use robots that combine short phrases from several words taken from thematic dictionaries, or change the words in the message for synonyms. In the end they get unique messages, making the task of spam filters more difficult. The small size of the emails also helps spammers to reduce traffic costs.

Malicious attachments in email

For the fourth year in a row the most widespread malware in emails were programs that attempted to steal confidential data, usually logins and passwords for Internet banking systems.

The Top 10 malicious programs spread by email in 2014

Trojan-Spy.HTML.Fraud.gen topped the rating again. It is generally distributed using phishing emails and is designed to look like an html page where users are invited to enter their confidential data.

Email-Worm.Win32.Bagle.gt is in second place. The main functionality of all email worms, including Bagle, is to collect electronic addresses from compromised computers and to send copies of itself to all email addresses found on an infected computer. Bagle email worms can also receive remote commands to integrate with other malicious applications.

#Spam levels have fallen consistently from a peak of 85.2% in 2009 to 66.76% in 2014 #KLReport

Tweet

Third came Trojan.JS.Redirector.adf, which was the most widespread malware in Q3. The malware spreads via email in a passwordless ZIP archive. It appears as an HTML page with an integrated script which, when opened by users, redirects them to an infected site. There it usually offers to load Binbot — a service for the automatic trading of binary options, which are currently popular on the net.

Representatives of the Bublic family occupy 4th and 7th positions in the Top 10. Their main functionality is the unauthorized download and installation of new versions of malware onto victim computers. They often download a ZeuS/Zbot modification. The Trojans of the Bublic family appear as EXE files but use the Adobe document icon to mislead the victim.

Email-Worm.Win32.Mydoom.l is in 5th place. This network worm with a backdoor functionality is spread as an email attachment via file sharing services and writable network resources. It harvests email addresses from infected computers so they can be used for further mass mailings. The worm also enables attackers to remotely control the infected computer.

In 2014, the proportion of spam in email traffic was 66.76% #KLReport

Tweet

Trojan-Banker.Win32.ChePro.ilc. ended the year in sixth position. This downloader appears as a CPL applet (a component of the control panel) and, as is typical for this type of malware, it downloads Trojans developed to steal bank information and passwords. These banking Trojans mainly target online customers of Brazilian and Portuguese banks.

Eighth isTrojan-Downloader.Win32.Dofoil.ea. This Trojan downloads other malicious programs onto the victim computer to steal various user information (mainly passwords) and send it to the fraudsters.

Backdoor.Win32.Androm.daxcame 9th. This malicious program belongs to the Andromeda/Gamarue family of universal bot modules. The main features of these malware programs are the ability to download, store and run executable files, downloading and loading DLL (without saving on disk), downloading plugins and the ability to update and delete themselves. The bot's functionality is enhanced with a system of plugins that can be downloaded by the cybercriminals whenever necessary.

Exploit.JS.CVE-2010-0188.f rounds off the Top 10. This particular exploit appears as a PDF file and uses a vulnerability in version 9.3 and lower of Adobe Reader. This vulnerability has been known for a long time and poses no danger to users who update their software regularly. However, when it encounters an old version of Adobe the exploit downloads and runs the executable file Trojan-Dropper.Win32.Agent.lcqs. The dropper installs and runs the malicious script Backdoor.JS.Agent.h, which collects information about the system, sends it to the attackers' server and receives various commands in response. The commands and the results of their execution are transmitted in an encrypted form.

The Andromeda family remains the most widespread malware family. It accounted for 11.49% of all malware detected in malicious attachments. These programs allow the attackers to secretly control infected computers, which often become part of a botnet.

Second came ZeuS/Zbot (9.52%), one of the most popular and widely-available programs designed to steal banking information and, as a consequence, users' money. This program is often downloaded on to a victim computer by loader programs distributed via spam mailings.

Bublik (8.53%) completes the Top 3. This is a family of malicious loader programs that downloads modifications of the Zeus/Zbot family onto a compromised computer.

Countries targeted by malicious mailshots

Distribution of email antivirus activations by country, 2014

For the third year in a row the Top 3 countries most targeted by malicious mailshots remains unchanged: the US, the UK and Germany. The USA (9.80%) maintained its leading position despite the 2.22 pp decrease in the number of antivirus activations. Britain came second with 9.63% (-1.63 percentage points). Germany was third with 9.22%.

Of special note is France (3.16%) which climbed from 16th to 9th position in the rating.

Russia (3.24%) occupied 8th place, one position up from the previous year.

Spammers' tricks

In 2014, spammers had a mix of old and new distribution tricks to lure in users.

We came across emails containing attached archives with the .arj extension. This format was introduced long ago and is rarely used now. Therefore, even users who are wary of attached archives do not always recognize this attachment as potentially dangerous. The ARJ archiver has a further advantage as it can reduce file sizes to the minimum.

In addition to nonstandard archives, spammers also sent out malicious emails containing files with unusual extensions for attachments, such as .scr. This extension usually denotes a screensaver.

One of the most common types of malicious spam and phishing are fake bank notifications. In 2014, spammers began to complicate the design of fake messages by adding more links to official resources and services from the organizations that were mimicked in the fake notifications. Obviously, the attackers hoped that an email with a few legitimate links would be recognized as legitimate by users and spam filters alike. Meanwhile, the email contained a single fraudulent link; after clicking it an archive containing a malicious program was downloaded onto the victim computer.

In some cases, cybercriminals used different URL shorteners to mask the real link. Eventually they redirected the user to a popular cloud storage where a malicious program was hosted under the disguise of an important document.

Phishing

When preparing statistics on phishing we applied the methodology that was first used in our report "Financial cyber threats in 2013" published in April 2014. As a result the data on phishing for 2014 should be compared with the data in that report (not with the report "Spam in 2013").

The data source

The report is based on the data about Antiphishing system activations collected by Kaspersky Security Network. The Antiphishing system contains of three components:

2 deterministic:

  • Offline phishing contains a database of the most relevant phishing wildcards* and is located on users' devices. It is triggered when the system encounters a link that matches one of the phishing wildcards in the database
  • Cloud anti-phishing contains all known phishing wildcards*. The system refers to the cloud if the user encounters a link that is not included in the local anti-phishing database. Cloud databases are updated much quicker than local databases.

Heuristic:

  • The heuristic web component of the antiphishing system. This component is triggered when a user clicks on a link to a page with phishing content but information about this page is not yet available in the Kaspersky Lab databases.

* Phishing wildcards are a set of symbols to describe a group of links detected by the system as phishing. One phishing wildcard can help detect several thousand active links to phishing sites.

In 2014 the computers of users of Kaspersky Lab products recorded 260,403,422 instances that triggered the antiphishing system. Of these, 55% (143,827, 512) involved activation of the deterministic component, and 45% (116, 575, 910) came from the heuristic web component.

Phishing links: not only in email

The deterministic components of the antiphishing system (cloud and offline) check links in the user's browser and messages received via IM or email. Only 6.4% of all the activations of these components come from links in emails. This suggests that today, instead of traditional phishing mass mailings, phishers are using other ways to spread links and new scams.

Kaspersky's #antiphishing system was triggered 260,403,422 times in 2014 #KLReport

Tweet

Currently links to phishing sites are more often distributed via social networks. It's not just about using stolen accounts; fraudsters are also involving unwitting users in sending out phishing links to their friends in social networks.

For example, in July 2014 this particular trick was used to spread a link to the petition in support of Uruguayan footballer Luis Suárez via social networks. To sign the petition, users had to enter their personal data, which then went to the phishers. After that, the victim was invited to share the link with his friends on Facebook. As a result, the link to the phishing page spread quickly among football fans and their friends.

An example of a phishing page distributed with the help of social networks

Phishing emails

All this should not be taken as evidence that fraudsters have stopped spreading phishing links via email. It is still the most popular way of distributing links to fake pages of financial institutions. Perhaps that's why scammers usually send out emails containing links to phishing sites on weekdays when the users check their e-mail \at work.

Activation of the deterministic components of the antiphishing system on the users' mail clients

Fraudulent schemes utilizing phishing emails with malicious attachments, HTML files or HTML forms inserted in the body of the message are also popular with fraudsters.

Attaching HTML files or HTML forms allows fraudsters to reduce the costs of maintaining the page on the Web. The standard scheme looks like this. User receive a fake notification from an organization that informs them their accounts have been blocked, their data has leaked or some suspicious activity was logged. To help solve the problem, users should update their personal information in the attached file or form. The data entered by the victims goes straight to the scammers. Over the past year we found dozens of mailings like this purporting to come from various financial institutions and other organizations.

An email with an attached HTML file

A phishing attack using HTML attachments to target the customers of a specific organization is often a means of extracting a wide range of financial information from victims, not of all of which necessarily relates to the company used in the scam.

An email with an attached HTML file

The examples above show the tricks that fraudsters use trying to get not only credentials to access online accounts, but also other personal information including bank card data.

Big game hunting

As mentioned above, when attacking customers of various organizations, fraudsters often try to get not only the victim's account data but also the bank card details and other confidential information. In this way, some scammers collect valid e-mail address, perhaps to sell them to spammers. Others pose as a bank or similar organization that is concerned about security and use that mask to steal the user's financial information and then his money.

The increased danger of phishing schemes consists of this: no matter what anti-fraud schemes are in place to protect customers (double, triple verification, one-time passwords, etc), they can only protect the account. If the user passes personal data to fraudsters there is little that can be done to stop that information being used to access the account.

We have already given examples of an "extended" phishing attack that utilizes HTML attachments. Yet another example is this phishing attack on PayPal. The traditional scam directs victims to a phishing page that mimics the site of the payment system. Unwary users input their usernames and passwords, sending them to the fraudsters. A further page opens and request bank card data and other information. The users believe they are safely logged onto PayPal and enter the information without a second thought. Once the scammers have that info, they redirect users to the official PayPal site leaving their victims none the wiser about the data theft.

A phishing attack on PayPal

The attackers may not be able to get into the victim's PayPal account because the company has additional protection measures. But they will have enough information to be able to steal money in other ways.

A phishing attack on one of the largest organizations in the telecommunications industry is another good example. At first glance, the fraudsters need a login and password to access the user's personal account. However, once the victim enters a fake account, the scammers not only ask for the card data but also for all other information that may be useful for them to manipulate the victim's money.

A phishing attack targeting the victim's personal information

The geography of attacks

In 2014 phishing attacks were registered in almost all countries worldwide.

Top 10 countries by percentage of attacked users

Brazil had the highest proportion of users subjected to phishing attacks (27.47%).

The percentage of users on whose computers the antiphishing system was triggered out of the total number of users of Kaspersky Lab products in the country, 2014

Top 10 countries by percentage of attacked users

  Country % of users 1 Brazil 27.45 2 Australia 23.76 3 India 23.08 4 France 22.92 5 Ecuador 22.82 6 Russia 22.61 7 Kazakhstan 22.18 8 Canada 21.78 9 Ukraine 20.11 10 Japan 19.51

In 2014, the percentage of attacked users in Brazil grew by 13.81 percentage points compared with 2013 (when Brazil was 23th in the rating). This intense burst of phishing activity in Brazil is probably connected to the 2014 World Cup, which brought thousands of fans from around the world to the South American country.

The distribution of attacks by country

Russia had the greatest share of phishing attacks, with 17.28% of the global total. The percentage of users in this country on whose computers the Kaspersky Lab antiphishing system was up 6.08pp from the previous year.

Distribution of phishing attacks by country in 2014

The increase in the number of attacks on the users in Russia is probably related to the deteriorating financial situation in the country in 2014. People are making more transactions as they try to invest their savings and make online purchases. At the same time many users are worried, giving scammers more opportunities to apply their social engineering techniques and play on those fears.

Last year's leader - the US (7.2%) - moved down to second place, with a 23.6 percentage point drop in the number of attacked users. It is followed by India (7.15%) and Brazil (7.03%), where rates increased by 3.7 pp and 5.11 pp respectively.

Organizations under attack

These statistics on the organizations used in phishing attacks are based on the triggering of the heuristic component of the antiphishing system. The heuristic component is triggered when a user follows a link to a phishing page and there is no information about this page in the Kaspersky Lab databases.

Distribution of organizations subject to phishing attacks by category, 2014

In 2014 we saw changes in the organizations targeted by phishers. Last year's leader, the 'Social networking and blogs' category dropped 19.62 pp to 15.77%, and was overtaken by 'Global portals', which gained 19.29 pp and reached 42.59%. Global Portals was previously described as 'E-mail' in earlier reports. The change is no big surprise: Google, Yahoo!, Yandex and similar companies are constantly developing their services and offering new options, from email and social networks to e-wallets. This is very convenient for users because everything is available under one account; it's also a boon for scammers because one password can unlock a huge range of digital resources. That's why Yahoo! and Google came into the top 3 most-frequently attacked organizations, sandwiching Facebook.

Top 3 attacked organizations

  Organisation % of phishing links 1 Yahoo! 23.3 2 Facebook 10.02 3 Google 8.73

In 2014 the numbers for Yahoo! (23.3%) grew by 13.3 percentage points compared with the previous year (partly due to a sharp increase in the number of fraudulent links to fake Yahoo! pages at the beginning of January 2014).

The share of phishing attacks on financial institutions was 28.74%, a 2.71 pp drop from 2013. Within this sector the percentage of attacks on the banking sector declined (down 13.79 pp from 2013). At the same time the numbers for "Online stores" and "Payment systems" rose by 4.78 pp and 9.19 pp respectively.

Distribution of financial phishing by the type of organization attacked, 2013

Distribution of financial phishing by the type of organization attacked, 2014

You can read more about financial phishing in our report "Financial cyber threats in 2014: times have changed".

Conclusion

The percentage of spam in email traffic continues to decline; we do not expect a significant change in these numbers in 2015.

We expect a further reduction in the amount of advertising spam and an increase in the number of fraudulent and malicious emails. At the same time we will see more carefully designed fake messages, in which attackers will use even more elaborated tricks (such as malicious attachments with unusual extensions like .arj, .scr).

Fraudsters use a variety of methods to distribute phishing content. However, time-proven phishing mass mailings are still popular and are likely to remain so for a long time.

For their phishing attacks scammers choose the clients of the most popular organizations, thereby increasing the likelihood of a successful attack. At the same time, many attacks are conducted in order to get maximum personal, primarily financial, information from the victim. We assume that this trend will continue in the future.

'Locked Out'

Thu, 03/12/2015 - 07:00

Today the great majority of malware is created with the aim of enrichment.  One of the tactics often used by evildoers is to encrypt files and demand a ransom for their decryption. Kaspersky Lab classes such programs as Trojan-Ransom malware, although there is another widely used and resonant name – encrypters.

Encrypters have become a serious problem for users, especially corporate users.  And related topics attract the most posts and readers on our forum.

Despite all the efforts of the anti-virus companies we don't expect an easy victory over encrypters in the short term.  There are at least two good reasons for this:

  1. Encrypters are constantly evolving.  It is a battle of arms and armour: the defence gets better – the weapons get better.
  2. The attack is not carried out on the user's computer but on the system of computer + user.  That is, one of the attack vectors is human.  A person is subject to emotions and irrational acts.  A person is capable of ignoring the warnings of the defence systems or turning it off altogether.  This is precisely what the evildoers are counting on.

In this article we look at the evolution of complication of the encryption schemes used by virus writers and the methods they adopt to put pressure on their victims.  At the end of the article there is some advice for users which might help them protect important files.

The evolution of encrypters: from simple to complex

Serious antivirus companies devote special attention to protection against encrypters. To counter the improved systems of defence virus writers need to change their programs regularly. And they change almost everything: the encryption schemes, means of obfuscation and even the formats of executable files.

Virus writers change the encryption schemes, means of obfuscation and even the formats of executable files

Tweet

We will consider the evolution of encrypters in terms of the methods of encryption and cypher schemes employed. Depending on the cypher scheme used and the method of obtaining the key, in some cases it is possible to easily decypher the encrypted data and in others it is impossilbe to do so within a reasonable time.

Encryption with an XOR operation

We begin with programs that use the most primitive encryption.  A typical example of such malware is the Trojan-Ransom.Win32.Xorist family.  It has the following characteristics:

  • Xorist is one of the few encrypters that carries out its threat and damages the users files when several incorrect attempts are made to enter the password.
  • An XOR operation is used to perform the encryption.  The vulnerability of this encryption scheme is that it is possible to easily decrypt files because of the well-known standard file headers.  To counter this attack Xorist encrypts files not from the very beginning but after an interval.  By default this interval is 104h bytes but this can be changed at compilation.
  • To complicate the encryption algorithm the key is randomised with the help of the first letter of the file name.

Fragment of a file encrypted by an encrypter of the Xorist family: the eight byte key is clearly visible

On the whole, despite all the cunning of the creators of Xorist the files encrypted by it can be entirely decrypted relatively easily.  Maybe for that reason at the moment the Xorist family of malware is hardly ever encountered in the wild.

To combat Trojan-Ransom.Win32.Xorist the specialists of Kaspersky Lab created the utility XoristDecryptor.

Symmetrical Encryption

A symmetrical encrytion scheme is a scheme that uses a pair of keys for encrytion and decryption that are symmetrical to each other (this is why this scheme is called symmetrical).  In the great majority of cases in such schemes one and the same key is used for encryption and decryption.

If the key is embedded in the body of the encrypter, if one has access to the body of the malware it is possible to extract the key and create an effective utility to decrypt the files.  Such malware usually tries to delete itself after encrypting the files.  An example of this type of program could be one of the modifications of the Rakhni family.  Keys that were detected were added to the utility RakhniDecryptor.

If the key is recieved from the attacker's server or generated and sent to it then having an example of the malware yields little — an example of the key is necessary, and it is on the attacker's server.  If it is possible to recover the key (for obvious reasons the malware tries to delete such key after use) then it is possible to create a utility for decryption.  In this case a system that caches the internet traffic of the user may be useful.  An example of this type of malware is Trojan-Ransom.Win32.Cryakl.

Assymetric encryption

Assymetric encryption is the name given to those schemes in which the encryption and decryption keys are not related in an obvious symmetrical way.  The encrytion key is called the open or public key and the decryption key is the secret or private key.  Calculating the private key from a known public key is a very complicated mathematical task which is not possible in a reasonable time using modern computing capabilities.

At the heart of assymetrical cypher schemes is the so-called trapdoor one-way function.  Put simply this is a mathematical function that depends on a parameter (secret).  Without knowing the secret parameter the value of the function is calculated comparatively easily going one way (for a given value argument we can calculate the value of the function) and extremely difficult in the reverse direction (knowing the value of the function to calculate the value of the argument).  However everything changes knowing the secret parameter — with its help it is possible to reverse the function without particular difficulty.

Assymetric encryption with one key pair

If the public key is embeded in the body of the malware the presence of the malware without the private key is almost no help in decyphering the files (but does help in detecting the program and others like it in the future).

However if the private key becomes known (and it should at least be contained in the decrypter which the evildoer is offering for sale), then it becomes possible to decrypt the data for all users affected by the modification of the program using that public key.

An example of malware of this type is Trojan-Ransom.Win32.Rector.  The characteristics of this family are as follows:

  • Uses assymetric encryption and the public key is hidden in the body of the encrypter.
  • To speed up the encryption of files it doesn't encrypt them all at once but in small sections.  The encrypted sections are added on to the end of the file and their space is filled in with sequences with a frequency of one byte.  Because of this the encrypted file gains a typical 'scratched' appearance.

File fragment encrpted by a program from the Rector family

  • One defect of this scheme for the evildoer is that for the decryption of the files it is necessary to hand over the private key, which can be used to decypher all files encrypted by this modification of the malware.

Thus, although direct decryption of the files is impossible, several users suffering from one and the same modification of the malware can unite and buy one decoder for all of them.  Also users and other interested persons send decoders to us.  The private codes received are added to the RectorDecryptor.

If the public key is obtained from the evildoer's server (which allows the use of a unique public key for each user) then the presence of the body of the malware doesn't help in the decryption of the data — it is necessary to have the private key.  However the body of the program helps identify and block the malware server and this helps protect other users.

Encryption using several keys

To ensure a unique decoder for each user schemes with several keys are used.  For this the key for encryption of data is generated on the victim's computer.  It might be a symmetric key or an assymetric key pair.  The algorithm for key generation is chosen so that the resulting key is unique for each affected user.  In other words the chances of these keys being the same in any two cases should be extremely small.  However sometimes the malware creators make a mistake and the key is generated from a relatively small range of possible values.  In this case the user's data can be decyphered by trying all possible values of the key.  However such cases have been rare lately.

The user's data is encrypted using the generated key.  Then the key that is necessary to decypher the data is encrypted itself using another public key.  This public key is generated earlier and the accompanying private key is not in the body of the encrypter but instead that private key is known to the evildoer.  Then the original key necessary for decyphering the data is deleted and only the encrypted version remains on the user's computer.

Now, having received the encrypted copy of the key the evildoer can extract the key from it that is needed to decypher the user's data and include it in the decoder.  And this decoder will be useless for other affected users.  Which, from the point of view of the evildoers, is a great improvement over the two-key schemes described above.

There is no algorithm to decrypt files encrypted with the RSA with a key length of 1024 bits in an acceptable time

Tweet

An example of malware using a scheme with several keys is the Trojan-Ransom.BAT.Scatter family. The Scatter family has several significant features:

  • A more advanced encryption scheme is used with two pairs of assymetric keys, which allows the evildoers to encrypt the files of the victim without revealing their private key.
  • Samples of this family are written in scripting languages, which allows the malicious functions to be easily changed.  Scripts are easier to obfuscate and this process is easier to automate.
  • The samples have a modular structure.  The modules are downloaded from the wrongdoers' website during the running of the script.
  • Renamed legitimate utilities are used for the encryption of files and deletion of the keys. 
  • A high level of automation of the process has been achieved.  Almost everything is automated, the malware objects are automatically generated, letters are sent out automatically.  Furthermore, according to the malefactors the process of handling letters from victims and further contact with the victims has been automated.  The decyphering of test files of the victim, evaluation of the cost of the information, the provision of bills, checking payment and sending out decoders all happen automatically.  It is difficult for us to check the truth of this information but taking into account data obtained from studying the modules of Trojan-Downloader.BAT.Scatter there is no reason not to believe these claims. 

The Scatter family appeared quite recently: the first samples were detected by Kaspersky Lab specialists at the end of July 2014.  In a short time it significantly evolved, providing itself with the functionality of Email-Worm and Trojan-PSW.

From 25 July 2014 to 25 January 2015 we detected 5989 attacks with the use of Trojan-Downloader.JS.Scatter on 3092 users.

Number of detected downloadings of Trojan-Downloader.JS.Scatter. The spike in the middle of November is the result of a new modification spreading in the USA

The geography of distribution of Trojan-Downloader.JS.Scatter downloads 25 July 2014 — 25 January 2015

This family is worth discussing in more detail as we can say with certainty that the Trojan-Downloader.*.Scatter family is a new step in the evolution of encrypters.

Technical details: Scatter, a new evolutionary step

The Scatter program family is multimodule script multifunction malware. As an example we chose the modification of the encryption module which is detected as Trojan-Ransom.BAT.Scatter.ab which started to appear with regularity in the middle of October.

More Trojan-Downloader.JS.Scatter.i download module

The malware download module is spread in email attachments. The filenames are specially chosen by the attackers to make the letter seem legitimate and end up with the accounting staff.

FullName HitsCount ./draft collation act.zip// unpaid bills. Draft collation act for two months – accountancy dept agreed till 14 October 2014_mail.attachment_scannеd.avast.ok.dос .js 4386 scan copy of debts 2014.zp//unpaid bills . Draft collation act for two months – accountanct dept agreed till 14 October 2014._mail.attachment_scannеd.avast.ok.dос .js 402 unpaid bills. Draft collation act for two months – accountancy dpet agreed till 14 October 2014_mail.attachment_scannеd.avast.ok.dос .js 241 Draft collation act.zip 22

The most popular names of the Scatter download modification appearing in the first half of October

If a user attempts to open the attachment they start the downloader, which is an obfuscated JavaScript and is detected by Kaspersky Lab as Trojan-Downloader.JS.Scatter.i

Fragment of the obfuscated code of the downloader Trojan-Downloader.JS.Scatter.i

After being started by the user the downloader downloads five other objects from the malefactor's site.  These files are saved in a directory defined by the variable %TEMP%.  Not all of these five objects are harmful:

  • fake.keybtc – is a renamed version of the legitimate program gnupg gpg.exe intended for carrying out cryptographic operations.
  • night.keybtc – is a renamed version of the library iconv.dll necessary for gpg.exe to work properly
  • trash.keybtc – is a renamed version of the utility sdelete.exe from Microsoft designed to reliably delete files.
  • key.block – is a malicious command script that uses the utilities above to encrypt files.  This object is detected by Kaspersky Lab as Trojan-Ransom.BAT.Scatter.ab
  • doc.keybtc – this file is in the Microsoft Word format.  The downloader renames this file as word.doc and then tries to run it.  If there is a program for looking at .doc files on the user's computer the user sees the following picture:

The beginning of the Microsoft Word document shown to the user by the downloader Trojan-Downloader.JS.Scatter.i

This document doesn't contain any malicious code.  Its task is too reduce the alertness of the user and distract his attention from the processes taking place on his/her computer.

In the meantime the downloader renames the file key.block to key.cmd and runs it.  At that the work of the downloader is finished and Trojan-Ransom.BAT.Scatter begins.

The sequence of actions of the encrypter Trojan-Ransom.BAT.Scatter.ab 1. Preparation
1.1. Rename the legitimate files it needs with extensions that can be used.
1.2. Check the presence of the special file containing in its name the client identifier and the current date. If such a file exists the encrypter considers that the files are already encrypted and doesn't do anything else. This prevents the rewriting of the special files KEY.PRIVATE and UNIQUE.PRIVATE, created by the Trojan during encryption (more details on these below).
1.3. Check the presence of the directory %AppData%\BitCoin. If this directory exists then later the Trojan tries to steal the BitCoin wallet data.
1.4. Check the existence of the file "%TEMP%\partner.id". This confirms the information found earlier about the presence of the partner programs spread by Scatter. (It is interesting that in some communications on infected computers the wrongdoers offered their victims to decypher their files in exchange for certain services and even promised money for these services. It is possible that in this way they are trying to turn the user into a partner.)
1.5. Generate a key pair (public and private keys: files pubring.gpg and secring.gpg respectively) with the parameters:
Key-Type: RSA
Key-Length: 1024

This type of encryption is currently considered effective: there is no algorithm to decrypt files encrypted with the algorithm RSA with a key length of 1024 bits in an acceptable time without knowing the private key.

1.6. Extract the public key from the body of the malware and  use it to encrypt the file secring.gpg, the private key of the key pair, as a result obtaining the file secring.gpg.gpg.  After that secring.gpg is deleted with the help of the legitimate utilitysdelete.exe and its location rewritten 16 times.  If for some reason it is impossible to delete the unencrypted key using sdelete the Trojan tries to delete it itself, writing over it several times with rubbish.  Multiple rewriting of the location of the file is necessary so that the private key can not be recovered even using special programs for restoring deleted data.
1.7. Copy the encrypted private key (secring.gpg.gpg) under the name %TEMP%\KEY.PRIVATE", which the malware tries to do twice for reliability.  Then it once more checks the presence of KEY.PRIVATE.  If it isn't there and neither is secring.gpg the Trojan doesn't carry out encryption and goes straight to distribution of its loader (item 3)
2. Encryption
2.1. Before the start of encryption the Trojan generates a script with a list of files which it will encrypt.  It does this in two stages:
  • First it looks for and adds to the file databin.lst the paths to files with the following extensions:
    *.xls *.xlsx *.doc *.docx *.cdr *.slddrw *.dwg *.pdf.
  • Then it adds to databin.lst the paths to files with the following extensions:
    *.mdb *.1cd *.accdb *.zip *.rar *.max *.cd *.jpg.

Why does it do this?  The RSA algorithm is reliable but extremely slow.  Therefore the malware 'is afraid' that it might start encrypting large files or a directory with a lot of photographs and that something might interfere with it.  For instance the user might switch off the computer.  Therefore the Trojan first of all tries to encrypt small files that are potentially important for the organisation and then moves on to media such as disks and other large volumes of data.

Apart from the list for encryption, the names of files and their size the database UNIQUE.BASE is added to the file.  This database contains the name of the computer and name of the user.  Later the database created will help the evildoers evaluate the size and value of the encrypted information, so as not to undersell their 'goods' and seek the maximum price for decryption.

Then the list of files and database are filtered from files located in utility directories.  As a result the 'filtered' files UNIQUE1.BASE and bitdata1.bin are created.

2.2. The file UNIQUE1.BASE is encrypted with the public key pubring.gpg, which was generated at the begining of the operation of the encrypter. The resulting encrypted file is renamed UNIQUE.PRIVATE and the file UNIQUE1.BASE is deleted.
2.3. The files UNIQUE.PRIVATE and KEY.PRIVATE are copied straightaway in several places so that the user can find them easily. These files are encrypted and the user can not decypher them without knowing the private key of the attackers.
2.4. The Trojan generates a message to the user and adds it to the autoloader:

Fragment of the message of the evildoers (translation from the Russian):

For system administrators:

1. Your information has been encrypted using RSA-1024 assymetric encryption, used by the military.  Breaking it is impossible.
During encryption the special ID-file KEY.PRIVATE was copied to various places on the computer.  Do not lose it!
For each computer a new ID-file is created.  It is unique and contains the code for decryption.  You will need this.
'Temporarily blocked' means that the files are modified on the byte-level using a public 1024 bit RSA key.

2. And so, our further actions are as follows:

2.1. You can contact us only using the email address ************@gmail.com
2.2. First of all you need a guarantee that we can decypher your files.
2.3. Contact us.  The structure of your email should be as follows:

  • include your ID-file KEY.PRIVATE (!!) - look for it on your computer, without it it will not be possible to re-establish your data.
  • 1-2 encrypted files to check the possibility of decryption
  • the approximate number of encrypted files/computers

2.4. You will recieve a guarantee and the cost of your key within one hour
2.5. Next payment should be made, the minimum cost will be 150 euros
2.6. We will send you your key, you should put it in the same directory as the decoder (DECODE.exe)
2.7. When the decoder is started the concealed decryption of your data is carried out.  You should not start this process more than once.
2.8. The process of decryption might take up to 12 hours in stealth mode.  At the end of the process the computer will reboot.

2.5. The Trojan renames bitdata1.bin (the script for the encryption of data generated earlier) as bitdata.cmd and starts it running. As a result the user's files are encrypted and the email address of the evildoers is added to their extensions.
2.6. After successful encryption the mark BITM is added to all files UNIQUE.PRIVATE and KEY.PRIVATE
3. Distribution of loader by electronic mail
3.1. The Trojan downloads additional components allowing it to collect passwords from the same site of the wrongdoers that the loader used earlier.  These components are downloaded in parts and assembled on the victim's computer.
3.2. With the help of the downloaded components the evildoer looks for user passwords for mail services Mail.ru, Yandex.ru and Gmail on the infected computer.  Any passwords found are sent to a special email address of the malefactor and data from any located BitCoin wallets are also sent there.
3.3. The malware generates 15 variants of letters.  They are all linked by a legal and not an accounting theme on this occasion.

With the help of passwords to mail services obtained earlier the Trojan connects with the mail servers and obtains the headers of letters received.  Email sendouts and automatic mesages are filtered out of the emails received.  All the remaining email addresses are sent one of the 15 possible versions of the letters, selected with the help of a random number generator.

It is interesting that regardless of the text of the letters, one and the same attachment is added — the archive with password '1'.  This archive is downloaded from the same site of the attacker before the start of the mail out.  Inside the archive is a file with a long name in Russian, which translates as:

Complaint concerning unpaid debts. Legal department — Confirmed and agreed for dispatch to debtor_October 2014_ Avast.ОК.dос .js

In several cases the theme of the letter and the name of the attachment do not match each other — this is a drawback of the automatic generation of letters and malware objects.

The object with the long name is the JavaScript Trojan-Downloader.JS.Scatter.i described earlier but already with another obfuscation.

Code fragment of the downloader Trojan-Downloader.JS.Scatter.i with another obfuscation

Despite the obfuscation both scripts are successfully detected by Kaspersky Lab products, both by signature and using heuristics written over a year ago, before the appearance of this type of malware.

To the aid of the bad guys: the human factor

The business of cyber-blackmailers is flourishing. In 2014 Kaspersky Lab recorded more than seven million attacks on its users with the use of objects from the Trojan-Ransom family.

In 2014 Kaspersky Lab recorded more than 7 million attacks with the use of encrypters

Tweet

Number of attacks by encrypters blocked every month by Kaspersky Lab in 2014

Malefactors ever more frequently prefer to receive payment in the crypto-currency BitCoin.  Although prices for users by habit are indicated in rubles, US dollars and euros.  The prices for decryption for simple users start at 1000 rubles and increase to several hundred dollars.  In the case of encryption of the files of an organisation the appetite of the malefactors increases by on average a factor of five.  There are cases known when 5000 euros was demanded for file decryption.  Unfortunately, for companies that have lost their data it is often simpler to pay than lose important information.  It is no surprise that organisations are the main target of evildoers utilising encrypters.

Why are encrypters able to inflict such damage?

As was mentioned above, most antivirus companies constantly improve their defences against encrypters.  For instance Kaspersky Lab has implemented special technical 'Protection against Encrypter Programs' in its products.  However, as is well known, the weakest point in IT protection is the user.  And in the case of encrypters this is extremely relevant.

We conduct special events dedicated to combatting this type of malware.  These events include a whole complex of measures: analysis of all incidents that have occured at organisations contacting our technical help service (using both our own and other antivirus products); search for and collection of samples of encrypters; analysis of the work of each defensive component of our products in each event that happened; improvement of existing and development of new methods of detecting and remedying the consequences of the actions of encrypters.  This is painstaking work and takes a lot of time, but it is necessary for our products to deal successfully with this constantly changing threat.

In our research we often see file encryption attacks made possible by employees working with antivirus disabled

Tweet

During these investigations we often come across instances of the encryption of files in organisations as a consequence of their employees working with the antivirus program switched off.  And these are not isolated cases, our technical help service encounters such cases several times a week.

It seems to us that one possible reason for such carelessness among users, strange as it may seem, is down to significant technical progress.  The improved defences of browswers and operating systems has led to a state where today users encounter the threats of malicous programs less often than previously.  As a result some of them, not thinking, switch off individual components of their antivirus products or don't use them at all.

Much has been said about the need to regularly update programs.  Nevertheless we once again note the importance of keeping anti-virus programs up to date.  We have investigated cases of encryption of files at organisations that happened for one simple reason: the user, on arriving at work, started to read their mail not waiting for the anti-virus database to update — and that update contained a signature capable of identifying the malware involved.

On the other hand it is worth remembering that no product, no matter how modern, can provide 100% protection against malware appearing on the computer.  Belief in the absolute defence of a 'super-anitvirus program' leads to users being careless — for instance opening file attachments in suspicious letters or unthinkingly clicking on dangerous links.  The availability of 'advanced' systems of defence does not relieve the user of the need to follow the security policy.

Make back-up copies of all important files on separate media off the computer

Tweet

The lack of back-up copies of important files plays its part in the success of encrypters.  Earlier it was possible to lose data not only as a result of the operation of malware but because of failure of the data medium or one's own legitimate programs, used to operate on important data.  But in recent decades the reliability of media and programs has improved dramatically.  And most users have stopped making back-up copies of their data.  As a result, if a computer is infected with an encrypter it simply paralyses the normal work of the company and the chances of the attacker receiving money for decrypting the data increase accordingly.

Traps for the unwary: how users are attacked

If you compile a hit parade of the methods used to spread encrypters the first and second places would be taken resoundingly by email. In the first case the dangerous object is contained directly in the letter and in the second the letter doesn't contain the object itself but a hyperlink to it. In third place in terms of popularity we see attacks via a system for remote control of the computer (Microsoft's Remote Desktop Protocol or RDP). Such attacks as a rule are carried out on an organisation's servers.

RDP attack

Let's start with the rarest and simplest method.  In the event of an RDP attack the evildoer, having obtained remote access to the computer, first of all switches off the antivirus program and then runs the encrypter.  The main factors allowing such an attack via RDP are the use of weak passwords or a leak of information about the password from the user's record files. The introduction of a strict password policy will help resist such an attack:

  • a password must be tough to crack (complicated);
  • a password should be known only to its user;
  • a password should be changed regularly.
Attack via electronic mail

If an attack by RDP occurs without the user's involvement; an attack via email must be activated by the user him or herself by running a received file or clicking on a link in a letter.  This is achieved by social engineering methods used by the wrongdoer or, to put it more simply, by lying to the user.  The wrongdoer's strategy is often built on the fact that the person under attack is chosen because they have a job totally unrelated to information security.  Such people may not even know of the existence of such threats as malicious encryption of files.

The person under attack is chosen because they have a job totally unrelated to information security

Tweet Letter topics

The organisation receives a letter that sounds frightening, for instance a court case has been initiated against the organisation, the details of which are contained in the document attached.

A example 'letter from the court'. The attachment contains a Trojan-encrypter

The thinking of the evildoers is probably something like the following: frighten the victim with some imaginary threat, the fear of which outweighs the worry about opening an unknown email attachment.

For organisations this approach works especially well: the simple employee receiving such a letter bears an unexpected responsibility.  The employee tries to share the responsibility and consults his/her colleagues.  The evildoer's chances  that someone will open the attachment increase.  In several incident investigations  it turned out that the in-house lawyers of the victim organisations insisted that the attachment be opened.

Be suspicious of links and attachments in unexpected letters

Tweet

And to reduce the suspicions of the recipient the author of the letter might use official logos:

An example of a letter containing a link to a malicious object

Or the executable file might be built into a Microsoft Word document and be masked by an icon:

An example of how an executable file can be hidden in a Microsoft Word document

The malefactors also use a scheme when a Microsoft Word document contains unreadable text and a request to allow macros, supposedly to correct the appearance of the text. In actual fact after the operation of the macro the Trojan-encrypter will be loaded onto the computer.

An example of a Microsoft Word document 'convincing' the user to execute a malicious macro
The red text says 'To correct the display switch on macros'

The thing about filenames

The next social engineering technique is the use of special words in the names of files contained in the archives attached to the letter (or downloaded by the user). For instance it could be the word 'checked' or 'secure' plus the name of various anti-virus products. The aim of the malefactors is to make the user believe that the attachment has been checked by an anti-virus product.

An example of a malicious attachment using the name of an anti-virus product and the extension .js

The extensions for executable files are specially chosen to be unknown to the casual user.  Usually .scr, .com and .js are used.

A special mention goes to attachments apparently providing 'free security tutorials from Kaspersky Lab'.  Such letters are also sent in the name of other security companies.

Recommendations for users

Detailed recommendations for system administrators can be found here.
Here we give some brief recommendations for users:

  • Make back-up copies of all important files on separate media off the computer.
  • Switch on display extensions for registered file types.  This will help you to check that the document sent to you really is a document and not an executable file.  You need to check this even if the letter comes from a known sender.
  • Be suspicious of links and attachments in unexpected letters.  Curiosity and fear are the favourite instruments of wrongdoers, causing users to forget about being cautious and to open attachments.
  • Use the latest version of anti-virus products. As a rule their effectiveness increases with every new version thanks to new modules.  We earnestly recommend the users of our products to enable KSN.
  • And finally, wait for the anti-virus database to be updated before reading your morning mail.
  • System administrators (in addition to everything else) should keep users aware of threats.

Inside the EquationDrug Espionage Platform

Wed, 03/11/2015 - 07:00
Introduction

EquationDrug is one of the main espionage platforms used by the Equation Group, a highly sophisticated threat actor that has been engaged in multiple CNE (computer network exploitation) operations dating back to 2001, and perhaps as early as 1996. (See full report here [PDF]).

EquationDrug, which is still in use, dates back to 2003, although the more modern GrayFish platform is being pushed to new victims.

EquationDrug represents the main espionage platform from the #EquationAPT Group

Tweet

It's important to note that EquationDrug is not just a Trojan, but a full espionage platform, which includes a framework for conducting cyberespionage activities by deploying specific modules on the machines of selected victims. The concept of a cyberespionage platform is neither new nor unique. Other threat actors known to use such sophisticated platforms include Regin and Epic Turla.

The EquationDrug platform can be extended through plugins (or modules). It is pre-built with a default set of plugins supporting a number of basic cyberespionage functions. These include common features such as file collection and the making of screenshots. Sophistication is added by storing stolen data inside a custom-encrypted virtual file system before it is sent to the command and control servers.

The name "EquationDrug" or "Equestre" was assigned to this framework by Kaspersky Lab researchers. The only reference left by the framework developers was a short string "UR", as seen in several string artifacts left in the binaries.

Platform Architecture

The EquationDrug platform includes dozens of executables, configurations and protected storage locations. Putting all the pieces of this puzzle together in the right order may take time for those who are not familiar with the platform.

The platform includes executables, configurations and protected storage locations #EquationAPT

Tweet

The architecture of the whole framework resembles a mini-operating system with kernel-mode and user-mode components carefully interacting with each other via a custom message-passing interface. The platform includes a set of drivers, a platform core (orchestrator) and a number of plugins. Every plugin has a unique ID and version number that defines a set of functions it can provide. Some of the plugins depend on others and might not work unless dependencies are resolved.

Similar to popular OS kernel designs, such as on Unix-based systems, some of the essential modules are statically linked to the platform core, while others are loaded on demand.

The hypothesis that these attackers have been active since the 90s seems realistic #EquationAPT

Tweet

The platform is started by the kernel mode driver component ("msndsrv.sys" on Windows 2000 or above and "mssvc32.vxd" on Windows 9x). The driver then waits for the system to start and initiates execution of the user-mode loader "mscfg32.exe". The loader then starts the platform's central module (an orchestrator) from the "mscfg32.dll" module. Additional drivers and libraries may be loaded by different components of the platform, either built-in or auxiliary.

Platform Components

The EquationDrug platform can be as sophisticated as a space station, but it appears to be of no use without its cyberespionage features. This function is provided by plugin modules that are part of the massive framework described above. We discovered dozens of plugins and each is a sophisticated element that can communicate with the core and become aware of the availability of other plugins.

The plugins we discovered probably represent just a fraction of the attackers' potential. Each plugin is assigned a unique plugin ID number (WORD), such as 0x8000, 0x8002, 0x8004, 0x8006, etc. All plugin IDs are even numbers and they all start from byte 0x80. The biggest plugin ID we have seen is 0x80CA. To date, we have found 30 unique plugin IDs in total. Considering the fact that the developers assigned plugin IDs incrementally, and assuming that other plugin IDs were assigned to modules that we have not yet discovered, it's not hard to calculate that 86 modules have yet to be discovered.

86 modules have yet to be discovered #EquationAPT

Tweet

The most interesting modules we have seen contain the following functionality:

  • Network traffic interception for stealing or re-routing.
  • Reverse DNS resolution (DNS PTR records).
  • Computer management:
    • Start/stop processes
    • Load drivers and libraries
    • Manage files and directories
  • System information gathering:
    • OS version
    • Computer name
    • User name
    • Locale
    • Keyboard layout
    • Timezone
    • Process list
  • Browsing network resources and enumerating and accessing shares.
  • WMI information gathering.
  • Collection of cached passwords.
  • Enumeration of processes and other system objects.
  • Monitoring LIVE user activity in web browsers.
  • Low-level NTFS filesystem access based on the popular Sleuthkit framework.
  • Monitoring removable storage drives.
  • Passive network backdoor (runs Equation shellcode from raw traffic).
  • HDD and SSD firmware manipulation.
  • Keylogging and clipboard monitoring.
  • Browser history, cached passwords and form auto-fill data collection.
Code Artifacts

During our research we paid attention to unique identifiers and codenames used by the developers in the malware. Most of this information is carefully protected with obfuscation or encryption algorithms to prevent quick recognition, but anyone who breaks through this layer of encryption may discover some interesting internal strings, as demonstrated below:

Some other interesting text strings include:

SkyhookChow Target
SkyhookChow Payload
Dissecorp
Manual/DRINKPARSLEY/2008-09-30/10:06:46.468-04:00
VTT/82053737/STRAITACID/2008-09-03/10:44:56.361-04:00
VTT/82051410/LUTEUSOBSTOS/2008-07-30/17:27:23.715-04:00
STRAITSHOOTER30.ex_
BACKSNARF_AB25
c:\users\rmgree5\co\standalonegrok_2.1.1.1\gk_driver\gk_sa_driver…
To install: run with no arguments
Attempting to drop
SFCriteria_Check failed!
SFDriver
Error detected! Uninstalling...
Timeout waiting for the "canInstallNow" event from the implant-specific EXE!
Trying to call privilege lib...
Hiding directory
Hiding plugin...
Merging plugin...
Merging old plugin key...
Couldn't reset canInstallNowEvent!
Performing UR-specific pre-install...
Work complete.
Merged transport manager state.
!!SFConfig!!

Some other names, such as kernel object and file names, abbreviations, resource code page and several generic messages, point to English-speaking developers. Due to the limited number of such text strings it's hard to tell reliably if the developers were native English speakers.

Link Timestamp Analysis

We have gathered a reasonably large number of executable samples to which we have been able to apply link timestamp analysis.

A link timestamp is a 4-bytes value stored in an executable file header. This value is automatically set by compiler software when a developer builds a new executable. The value contains a detailed timestamp including minutes and even seconds of compilation time (think of it as the file's moment of birth).

Link timestamp analysis require the collection of the timestamps of all available executables, grouping them according to certain criteria, such as the hour or day of the week, and putting them on a chart. Below are some charts built using this approach.


Can we trust this information? The answer is: not fully, because the link timestamp can be altered by the developer in a way that's not always possible to spot. However, certain indicators such as matching the year on the timestamp with the support of technology popular in that year leads  us to believe that the timestamps were, at the very least, not wholly replaced. Looking at this from the other side, the easiest option for the developer is to wipe the timestamp completely, replacing it with zeroes. This was not found in the case of EquationDrug. In fact, the timestamps look very realistic and match the working days and hours of a well-organized software developer from timezone UTC-3 or UTC-4, if you assume that they come to work at 8 or 9 am.

The timestamps match the working days of software developer from timezone UTC-3 or UTC-4 #EquationAPT

Tweet

And finally, in case you are wondering if the developers work on public holidays, you can check this for yourself against the full list of their working dates:

2001.08.17 2007.12.11 2009.04.16 2011.10.20 2012.08.31 2013.06.11 2001.08.23 2007.12.17 2009.06.05 2011.10.26 2012.09.28 2013.06.26 2003.08.16 2008.01.01 2009.12.15 2012.03.06 2012.10.23 2013.08.09 2003.08.17 2008.01.23 2010.01.22 2012.03.22 2012.11.02 2013.08.28 2005.03.16 2008.01.24 2010.02.19 2012.04.03 2012.11.06 2013.10.16 2005.09.08 2008.01.29 2010.02.22 2012.04.04 2013.01.08 2013.11.04 2006.06.15 2008.01.30 2010.03.27 2012.04.05 2013.02.07 2013.11.26 2006.09.18 2008.04.24 2010.06.15 2012.04.12 2013.02.21 2013.12.04 2006.10.04 2008.05.07 2011.02.09 2012.07.02 2013.02.22 2013.12.05 2006.10.16 2008.05.09 2011.02.23 2012.07.09 2013.02.27 2013.12.13 2007.07.12 2008.06.17 2011.08.08 2012.07.17 2013.04.16   2007.10.02 2008.09.17 2011.08.30 2012.08.02 2013.05.08   2007.10.16 2008.09.24 2011.09.02 2012.08.03 2013.05.14   2007.12.10 2008.12.05 2011.10.04 2012.08.14 2013.05.24   Conclusions

EquationDrug represents the main espionage platform from the Equation Group. It's been in use for over 10 years, replacing EquationLaser until it was replaced itself by the even more sophisticated GrayFish platform.

The EquationDrug case demonstrates an interesting trend: a growth in code sophistication #EquationAPT

Tweet

The EquationDrug case demonstrates an interesting trend that we have been seeing while analyzing supposedly nation-state cyberattack tools: a growth in code sophistication. It is clear that nation-state attackers are looking for better stability, invisibility, reliability and universality in their cyberespionage tools. You can make a basic browser password-stealer or a sniffer within days.  However, nation-states are focused on creating frameworks for wrapping such code into something that can be customized on live systems and provide a reliable way to store all components and data in encrypted  form, inaccessible to normal users. While traditional cybercriminals mass-distribute emails with malicious attachments or infect websites on a large scale, nation-states create automatic systems infecting only selected users. While traditional cybercriminals typically reuse one malicious file for all victims, nation-states prepare malware unique to each victim and even implement restrictions preventing decryption and execution outside of the target computer.

Nation-state attackers create automatic systems infecting only selected users #EquationAPT

Tweet

Sophistication of the framework is what makes this type of actor different from traditional cybercriminals, who prefer to focus on payload and malware capabilities such as implementing a long list of custom third-party software credential database parsers.

The difference in tactics between cybercriminals and nation-state attackers appears to be due to relative resource availability. It's known that cybercriminals attempt to infect as many users as possible and that they can sometimes compromise hundreds of thousands of systems. It would will take many years to check all those machines manually, analyzing who owns them, what data is stored on them, and what custom software they run.

Cybercriminals probably don't even have enough disk space to collect all the potentially interesting data from the victims hit by their large scale infections. That is why cybercriminals prefer to extract tiny chunks of the most important data (credentials, credit card numbers, etc) on the machine of the victim and transfer only few kilobytes from each compromised host. Such data, when combined from all users, normally takes up gigabytes of disk space.

Nation-state attackers have sufficient resources to store as much data as they want. They have access to virtually unlimited data storage. However, they don't need, and often try to avoid, infecting random users, for the obvious reason of avoiding attention and remaining invisible. Implementing custom data format parsers in the malware not only doesn't help them find all the valuable data on the victim's machine, but may also attract extra attention from security software running on the system. They mostly prefer to have a generic remote system management tool that can copy any information they might need even if it causes some redundancy. However, copying large volumes of information might slow down network connection and attract attention, especially in some countries with poorly developed internet infrastructure. To date, nation-state attackers have had to balance between these two poles: copying victims' entire hard drives while stealing only tiny bits of passwords and keys.

Nation-state attackers use a remote system management tool that can copy any information they need #EquationAPT

Tweet

Now, if you wonder why EquationDrug, a powerful cyberespionage platform, doesn't include Skype or ICQ password stealing-capability, the answer is that they most likely copy the database as a whole and parse on the server-side. We believe that this will become a unique trademark of nation-state attackers in the future.

Some code paths in EquationDrug modules lead to OS version checks including a test for Windows 95, which is accepted as one of supported platforms. While some other checks will not pass on Windows 95, the presence of this code means that this OS was supported in some earlier variants of the malware. Considering this and the existence of components designed to run on Windows 9x (such as VXD-files), as well as compilation timestamps dating back to early 2000s, the hypothesis  that these attackers have been active since the 90s seems realistic. This makes the current attacker an outstanding actor operating longer than any other in the field.

Technical Details Kernel mode stage 0 (Windows 9x) - mssvc32.vxd MD5 0a5e9b15014733ee7685d8c8be81fb0d Size 6 710 bytes Format Linear Executable (LE)

This VXD driver handles only two control messages: W32_DeviceIoControl and Dynamic_Init. The DeviceIoControl part is not completely implemented and the driver is only able to check for some known control codes.  However it does nothing. This handler looks more like a code stub rather than actual payload.

On the Dynamic_Init event, the driver retrieves the location of the user-mode loader executable from the following registry value:

[HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] Config

If the value is not present in the registry, it uses the following fallback string hardcoded in the binary:

C:\WINDOWS\SYSTEM\SVCHOST32.EXE

Next, it installs a callback procedure using Windows function _SHELL_CallAtAppyTime. This procedure will be called when CPU is running in ring-3 mode, so that a new executable (loader process) can be started via the traditional way. This is a standard trick that was used by developers in the 90s to initiate a call to DLL export in ring-3 from ring-0 in Windows 9x OS family.

Kernel mode stage 0 and rootkit (Windows 2000 and above) - msndsrv.sys MD5 c4f8671c1f00dab30f5f88d684af1927 Size 105 392 bytes Format PE32 Native Compiled 2008.01.23 14:12:33 (GMT) Location %System32%\drivers\msndsrv.sys

This module can create log files in the following known locations:

%systemroot%\system32\mslog32.dat
%systemroot%\system32\msperf32.dat (default location)

The driver acts as the first stage of the EquationDrug platform on Windows 2000+ and implements rootkit functions for hiding the components of the platform. Additionally, it implements a NDIS driver for filtering network traffic.

When started and initialized, the driver retrieves the location of the user-mode loader executable from the registry value:

[HKLM\System\CurrentControlSet\Services\%driver name%] Config

The %driver name% is not hardcoded and is obtained dynamically from the current module name, which means that different instances may check different registry keys and this may not be a reliable way to check for infection. The sample we analyzed used "msndsrv" as the %driver name%.

Next, it crafts and injects a shellcode in "services.exe" or "winlogon.exe". The shellcode is designed to spawn the loader process from the executable called "mscfg32.exe".

The rootkit code in the driver hooks several Native API functions that lets it hide or protect registry keys, files and running processes. The components of EquationDrug can modify the list of protected objects by sending DeviceIoControl messages to the driver. The driver also maintains a persistent list of protected objects that is stored in the following registry values:

[HKLM\System\CurrentControlSet\Services\%driver name%] 1
[HKLM\System\CurrentControlSet\Services\%driver name%] 2

These values are also protected by the rootkit. They can be revealed by booting Windows in Safe Mode.

The driver contains the following unused strings:

  • \\.\mailslot\dskInfo
  • Dissecorp
User-mode loader - mscfg32.exe, svchost32.exe MD5 c3af66b9ce29efe5ee34e87b6e136e3a Size 22 016 bytes Format PE32 EXE Compiled 2008.01.23 14:26:05 (GMT) Location %System32%\mscfg32.exe

This module opens a unique event named "D0385CB7-B834-45d1-A501-1A1700E6C34E". If the event exists, it waits for 10 seconds and attempts to open a file whose name can be decrypted as "\\.\MSNDSRV". If the device file is successfully opened, the code issues a device request with IOCTL code 0x80000194 and no parameters.

This module uses RC5 in CBC-like mode with a key length of 96-bit for string encryption.

Careful analysis reveals some bits of uninitialized memory found next to encryption key locations. This is unused but partly meaningful memory, because it seems to contain short chunks of strings resembling some local filepaths:

  • "rver\8" (probably part of "Server\8..." string)
  • "LInj" (could be a part of "DLLInjector" or similar)

It's apparent that some parts of the code were designed to run on Windows 9x, for example a call to RegisterServiceProcess Windows API function makes sense only on Windows 9x OS family, because this API function doesn't exist on Windows NT platform.

The module uses a unique algorithm for generating registry value names. The code contains strings, such as "SkyhookChow Target", that are converted to GUID-like strings by calculating SHA1 hash and using its hexadecimal representation as a string. The resulting strings are used as actual registry value names in [HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] registry key.

Sample registry value names:

Original String GUID-like registry value name SkyhookChow Target {B6F5CD13-A74D-8B82-A6AA-6FA1BE2484C1-6832DF06} SkyhookChow Payload {F4CF0326-6DCD-EEC8-5323-01CEDB66741A-B55F6F12}

These registry values are encrypted using an RC5 algorithm using a hardcoded 1024-bit key with 24 rounds.

The registry value:

[HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] {F4CF0326-6DCD-EEC8-5323-01CEDB66741A-B55F6F12} ("SkyhookChow Payload")
should contain the location of the orchestrator DLL file ("mscfg32.dll"). If the value is not present a default value "%SYSTEM%\mscfg32.dll" is used.

The registry value:

[HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] {B6F5CD13-A74D-8B82-A6AA-6FA1BE2484C1-6832DF06} ("SkyhookChow Target")
may contain the location of the executable file that will be used as a "shell" process for the orchestrator library.

The module attempts to start the "shell" process in suspended mode. If there is no "SkyhookChow Target" value or the specified executable fails to start, the module tries different failsafe locations of the programs that can be used instead:

  1. Default browser set in the registry [HKLM\SOFTWARE\Clients\StartMenuInternet\{current @default value}\shell\open\command]
  2. %SystemRoot%\System32\svchost.exe
  3. %SystemRoot%\System32\lsass.exe
  4. Spoolsv service binary from the [HKLM\SYSTEM\CurrentControlSet\Services\Spooler] ImagePath registry value.
  5. Default html file handler from [HKLM\SOFTWARE\Classes\htmlfile\shell\open\command]registry value.
  6. Internet Explorer path from [HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths\] IEXPLORE.EXE registry value.

Next, the module injects extra code into a newly started target process. The injected code loads the payload DLL ("mscfg32.dll") into the target process and waits for the parent process to exit. When the parent process quits, it unloads the payload DLL and exits as well. The rest of the logic relies on the loaded DLL in that new process. See the description of the "mscfg32.dll" module below.

The module communicates with the Stage0/Rootkit driver "msndsrv.sys" by sending DeviceIoControl messages to the device "\\.\MSNDSRV". It activates the rootkit for its own process, for the target process holding the orchestrator and for all the files involved.

Platform orchestrator - mscfg32.dll, svchost32.dll MD5 5767b9d851d0c24e13eca1bfd16ea424 Size 249 856 bytes Format PE32 DLL Compiled 2008.01.24 22:11:34 (GMT) Location %System%\mscfg32.dll

Creates mutex: "01C482BA-BD31-4874-A08B-A93EA5BCE511", or terminates if one already exists.

Writes a timestamped log file to one of the following locations:

  • %SystemRoot%\temp\~yh56816.tmp
  • C:\Windows\Temp\~yh56816.tmp
  • %Registry_SystemRoot_Value%\temp\~yh56816.tmp
  • Value of [HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] D

The file "~yh56816.tmp" retains the history of execution. It comprises debug records of simple structure:

        Stage: DWORD | DateTimeLow: DWORD | DateTimeHigh: DWORD

Basically, it logs the execution of every stage of the orchestrator and the time of execution. The Stage is an integer number starting from 1.

This module spawns a new thread in the DllMain function which contains the main function body. The procedure disables application error popups shown by the default exception handler. This is probably done only in the "Release" version of the malware, because the following code generates exceptions that are reported to the user if application error popups are not disabled. We assume that the "Debug" version of the code doesn't suppress error popups when exception occurs as this helps with the debugging of the code.

The module checks the OS version and if it encounters an unsupported operating system the code generates an exception which terminates the application. The list of OS versions that pass this test:

  • Windows 95/98/ME
  • Windows NT 4.0 and above.

If the module runs on Win9x, it executes Win9x-specific function RegisterServiceProcess to hide from the Windows Task Manager application. If the module is NOT running on WinNT6.0+, it then attempts to open a virtual device file with one of the following names:

  • \\.\MSSVC32 on Win9x
  • \\.\MSNDSRV on WinNT

If the device file is successfully opened, the module activates a rootkit for its process and for the file location "%SYSTEM%\unilay.dll" local path. This is followed by finding and terminating a process named "winproc.exe" which is the name of another component of the platform. Note that this part of the code is executed only on platforms different from WinNT 6.x (Windows Vista and later).

The module was designed to fetch or update its main configuration data from different places. There are some default values set inside the code, such as some timeout values and the following C&Cs:

  • www.waeservices[.]com
  • 213.198.79.49

These default values can be overwritten later.

Next, it locates a data section called "Share2" in the current module and verifies the starting magic number. If it is 0x63959700, it then decrypts the rest of the data in the section and interprets it as a configuration block. However, data from the next location can override all previous settings. This is a registry value with special name.

The naming of the registry location is the same GUID-like SHA1 value as the one used in the loader ("mscfg32.exe"), and is produced from the source string "Configuration":

[HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] {42E14DD3-F07A-78F1-7659-26AE141569AC-E0B3EE89}

The configuration block stored in the registry value is encrypted using RC5 with the 1024-bit key. Both the loader and the orchestrator share the same key for encrypting and decrypting the registry values in the "MemSubSys" key.

The decrypted configuration block consists of a series of tagged configuration records in the following format:

        [RecordType:DWORD][RecordSize: DWORD][RecordValue: %RecordSize%]

We retrieved a copy of a configuration block and decrypted and partly interpreted it. We are including the results for one of the configuration blocks:

Time value: 1 year 0 months 1 days 22 hours 6 mins 52 secs. The orchestrator is expected to set this field to the time of initial configuration.
Binaries: 3x1024-bit encryption keys
1b8e7818dad6345c53c2707a2c44648eee700d5cf34fea6a19a3fa0a6a871c72963fdded 91e2703c82b7747b8793e3063700da32cfb8d907dcce1beb36edd575418d1134ef188b 27ec3ce23711a656b0a8bf28921fbf1c39b4c90ad561e4174ed90f26ce11245bb9deb4b 4720403f47ca865ec8bbd3c1df9d93d042ff5b52ec6
05000000000000000000000000000000000000000000000000000000000000000000 00000000000000000000000000000000000000000000000000000000000000000000 00000000000000000000000000000000000000000000000000000000000000000000 0000000000000000000000000000000000000000000000000000
ed04953f3452068ae6439f04c7904c8be5e98e66e2cd0f267d65240aeed88bd4d3c6105 c99950dd42ccde4bc6bbaf9f6cb1b4e628d943e91f8f97f2aff705fdd25e3af6ba0bc4fd13 d67a2bcb751bb8f21f3d4b66c599f3e572802911394d142f8cf3a299d6d4558f9f0f01634 9afd1888472f4f8c729ffe913f670931f1a227
C&C domain: www[dot]waeservices[dot]com
C&C IP address: 213.198.79.49
C&C port: 443
Timestamp: 2010-12-08 11:35:57
Tool Reference: VTT/82055898/STEALTHFIGHTER/ 2008-10-16/14:59:06.229-04:00
TimeoutA: 25200 sec (7 hours)
TimeoutB: 32400 sec (9 hours)
TimeoutC: 3600 sec (1 hour)
TimeoutD: 172800 sec (48 hours)
+Several Unknown Values

Other configuration blocks we discovered contained similar information, with only some unique values:

Timestamp: 2009-11-23 14:10:15
Tool Reference: Manual/DRINKPARSLEY/2008-09-30/10:06:46.468-04:00
Tool Reference: VTT/82053737/STRAITACID/2008-09-03/10:44:56.361-04:00
Tool Reference: STRAITSHOOTER30.ex_
Tool Reference: VTT/82051410/LUTEUSOBSTOS/2008-07-30/17:27:23.715-04:00
Tool Reference: BACKSNARF_AB25

During the next step, the module obtains PE file version information from the resource section. It loads the version info using hard-coded module names, which are supposed to match the current module name:

  • SVCHOST32.DLL for Windows 9x
  • MSCFG32.DLL for Windows NT

If file version information is available, it gets language-specific values of the PrivateBuild block. The codepage and languages that are verified: Unicode, LANG_NEUTRAL and LANG_ENGLISH_US. When this check passes, the module gets @default registry value from the following location:

  • [HKLM\SOFTWARE\Classes\CLSID\{091FD378-422D-A36E-8487-83B57ADD2109}] TypeLib

If the key is not found, the code checks for registry value TypeLib in the following key:

  • [HKLM\SOFTWARE\Classes\CLSID\{091FD378-422D-A36E-8487-83B57ADD2109}]

If such a value is found, it is then deleted along with the Version value if it exists in the same key.

The string obtained from one of two possible registry values is processed as if this value is a CLSID-like string: the code takes the last 16 hexadecimal digits, splits them in two 8-chars values, converts them to binary form (two DWORDs) and reverses the order of bytes in each DWORD and XORs, the first value with 0x8ED400C0, and the second with 0x4FC2C17B.  Next, the first DWORD value becomes second and the second becomes first. In this order, they are stored in a structure in memory. These two values seem to be very important as they override a few values in the previously known configuration. If they don't exist, values from the current configuration replace them and are stored back in the registry following the reverse procedure:

  1. [HKLM\SOFTWARE\Classes\CLSID\{091FD378-422D-A36E-8487-83B57ADD2109}\Version] is created and @default value is set to version obtained from file version information PrivateBuild field (i.e. 3.04.00.0001). This seems to be used as kit version number.
  2. [HKLM\SOFTWARE\Classes\CLSID\{091FD378-422D-A36E-8487-83B57ADD2109}\Version] is created and @default value is set to a CLSID like string generated from the following:
    • Fixed prefix string: "{8C936AF9-243D-11D0-"
    • Two important DWORD values in the format of "%04X-%04X%08X}" string.

We collected and decrypted several samples of such values. According to the code, they are initialized with values of the Microsoft filetime format. So, we decided to interpret them as filetime values:

20101C04EC2C17B: 1 year(s) 7 month(s) 21 day(s) 23 hour(s) 32 min(s) 1 sec(s)
81E01C04EC2C17B: 1 year(s) 7 month(s) 8 day(s) 12 hour(s) 13 min(s) 5 sec(s)
E0001C04EC2C17B: 1 year(s) 7 month(s) 21 day(s) 1 hour(s) 6 min(s) 15 sec(s)
77101C04EC2C17B: 1 year(s) 5 month(s) 20 day(s) 19 hour(s) 15 min(s) 4 sec(s)
30F01C04EC2C17B: 1 year(s) 8 month(s) 0 day(s) 6 hour(s) 10 min(s) 33 sec(s)
C0901C04EC2C17B: 1 year(s) 8 month(s) 2 day(s) 6 hour(s) 29 min(s) 39 sec(s)
66701C04EC2C17B: 1 year(s) 6 month(s) 9 day(s) 2 hour(s) 10 min(s) 23 sec(s)
F6501C04EC2C17B: 1 year(s) 6 month(s) 6 day(s) 19 hour(s) 53 min(s) 22 sec(s)
01401C04EC2C17B: 1 year(s) 6 month(s) 25 day(s) 23 hour(s) 34 min(s) 13 sec(s)

After that, the module stores current time values in encrypted form in the registry value:

[HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] {08DAB849-0E1E-A1F0-DCF1-457081E091DB-117DB663} (encoded SHA1 of "StartTime")

The module contains an additional compressed Windows DLL file in the resource section, which is extracted as "unilay.dll" (see below). This DLL exports a number of functions that are just wrappers of the system API used to work with files and the registry, and also start processes and load additional DLL files.

The orchestrator contains several built-in plugins that form the core of the platform. These are initialized in the first place, and then additional plugins are loaded. All the plugins are indexed in a single encrypted registry value:

[HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\MemSubSys] 1

This value has information about all the components of the current kit. It may include Unicode strings with paths to extra DLLs which serve as plugins. Each DLL exports at least four functions which are imported by ordinal numbers from 1 to 4.

The structure of the registry value "1":

[Count:DWORD]{ [Plugin Id:WORD][Plugin Path Length:DWORD][Plugin Path String:VARIABLE] }

Plugins interact with each other and with the orchestrator by exchanging messages of pre-defined format. The message transport is implemented as a global object that contains four communication streams. Every stream contains a pair of kernel synchronization object handles (a semaphore with fixed maximum value defaulted to 1000 and a mutex) and a message queue as an array. A dedicated thread processes messages that appear in the message queues.

A message arrives in a parcel, represented as two DWORD values that contain the size of the message and a pointer to the message data. The message data starts with a DWORD identifying a class of message (a request, reply, etc).

The orchestrator contains the following built-in plugins (listed by internal ID): 8000, 8022, 8024, 803C, 8046, 800A, 8042, 8002, 8004, 8006, 8008, 8070, 808E. Several additional built-in modules have been discovered in newer versions of the orchestrator that was shipped with the GrayFish platform.

EquationDrug Plugins: Plugin ID File name Description 8000 Built-in Core, basic API for other modules 8002 wshcom.dll C&C communication using Windows sockets 8004 Built-in Additional message queue 8006 Built-in Memory allocation / storage 8008 vnetapi32.dll& C&C communication code based on DoubleFantasy, using WinInet API 800A Built-in C&C communication orchestrator 800C perfcom.dll HTTP communication 8022 khlp680w.dll System API: execute processes, load libraries, manipulate files and directories 8024 cmib158w.dll Collects system information: OS version, computer name, user name, locale, keyboard layout, timezone, process lists 8034 cmib456w.dll Management of the VFS backed by encrypted ".FON" files in the "Fonts\Extension" directory. Provides encryption using RC5 for these files 803E nls_874w.dll Network sniffer 803C Built-in Communication with the NDIS filter part of "msndsrv.sys" 8040 khlp807w.dll Network exploration API, share enumeration and access 8042 Built-in Compression library based on Nrv2d / UCL 8046 Built-in Communication with the rootkit part of "msndsrv.sys" 8048 mstkpr.dll Disk forensics and direct NTFS reader based on sources of SleuthKit 8050 khlp760w.dll Additional encryption facilities for the file-backed VFS 8058 khlp733w.dll Collects local system information, WMI information, cached passwords 8070 khlp747w.dll Enumerates processes and system objects 807A mscoreep32.dll Plugins for monitoring Internet Explorer and Mozilla browser activities 808A khlp866w.dll Compression library based on Zlib 808E Built-in Reverse (PTR record) DNS resolver 8094 Built-in In-memory storage 809C Built-in In-memory storage 80AA nls933w.dll HDD / SSD firmware manipulation 80AE wpl913h.dll Keylogger and clipboard monitoring (aka "GROK") 80BE vnetapi.dll C&C communication via WinHTTP API 80C6 webmgr.dll Extracts web history, Mozilla/Internet Explorer-saved form data and cached credentials 80CA wshapi.dll C&C communications interface via Windows sockets Additional components Unilay.DLL

This module provides a compatibility layer for accessing system API functions for Windows 9x. It redirects Unicode ("W") variants of Windows API functions to corresponding ANSI variants by converting Unicode string parameters to multi-byte strings and calling the respective ANSI API.

MD5 EF4405930E6071AE1F7F6FA7D4F3397D Size 9 728 bytes Compiled 2008.01.23 14:23:10 (GMT) Format PE32 DLL, linker version 6.0 (Microsoft Visual C++ 6.0)

Exported functions (redirected to ANSI variants):

  • 100017EF: CopyFileW
  • 10001039: CreateDirectoryW
  • 10001111: CreateFileW
  • 100011B3: CreateProcessW
  • 10001177: DeleteFileW
  • 10001516: FindFirstChangeNotificationW
  • 10001466: FindFirstFileExW
  • 10001300: FindFirstFileW
  • 100014C6: FindNextFileW
  • 10001564: GetCurrentDirectoryW
  • 1000188F: GetFileAttributesW
  • 100016C6: GetStartupInfoW
  • 10001602: GetSystemDirectoryW
  • 10001664: GetWindowsDirectoryW
  • 10001853: LoadLibraryW
  • 1000178B: MoveFileExW
  • 1000172D: MoveFileW
  • 10001913: RegCreateKeyExW
  • 100019F5: RegDeleteKeyW
  • 10001DDF: RegDeleteValueW
  • 10001A39: RegEnumKeyExW
  • 10001BE2: RegEnumValueW
  • 1000199B: RegOpenKeyExW
  • 10001B23: RegQueryInfoKeyW
  • 10001D57: RegSetValueExW
  • 100010D5: RemoveDirectoryW
  • 10001E81: SHGetFileInfoW
  • 100015C6: SetCurrentDirectoryW
  • 100018CB: SetFileAttributesW
  • 10001E23: lstrcmpW
Network-sniffer/patcher - atmdkdrv.sys MD5s 8d87a1845122bf090b3d8656dc9d60a8
214f7a2c95bdc265888fbcd24e3587da Size 41 440, 43 840 bytes Format PE32 Native Compiled 2009.04.16 17:19:30 (GMT)
2008.05.07 19:55:14 (GMT) Version Info
  • FileDescription: Network Services
  • LegalCopyright: Copyright (C) Microsoft Corp. 1981-2000
  • InternalName: atmdkdrv.sys

or

  • FileDescription: CineMaster C 1.1 WDM Main Driver
  • LegalCopyright: Copyright 1999 RAVISENT Technologies Inc.
  • InternalName: ATMDKDRV.SYS

Creates a file storage "\SystemRoot\fonts\vgafixa1.fon". Its first word is set to 0x21 at the beginning of the DriverEntry function, and is replaced with 0x20 at the end of DriverEntry.

This driver appears to have been put together in "quick-and-dirty hack" style, using parts of the "mstcp32.sys" sniffer and other unknown drivers. It contains a lot of unused code which is partially broken or disabled. These include a broken "Dynamically disable/enable windows audit logging" subsystem and an incomplete "Patcher mode".

There are three algorithms used for strings encryption - RC5; alphabet encryption like the one used in "mstcp32.sys"; and XOR with a pre-seeded random number generator. Decrypted strings are immediately encrypted back until the next usage to avoid in-memory detection.

The driver's filename and device name differ across the samples. They depend on the name of the registry key that is used to start the driver.

The driver may operate in one of two independent modes - as a network sniffer or as a memory patcher. The mode of operation is selected on startup, based on the "Config2" value of the driver's registry key. By default the driver starts in "sniffer mode".

Sniffer mode

The sniffer code is similar to the one used in the driver's "tdip.sys" and "mstcp32.sys" and uses NT4 NDIS-4, XP NDIS-5 interfaces, targeting incoming traffic on Ethernet and VPN (ndiswanip) interfaces. It captures only directed packets (containing a destination address equal to the station address of the NIC). Packers-filtering engine rules may be set via DeviceIoControl messages. Filtered packets are stored in-memory until requested. Maximum packets storage list length is 128 items per filtering rule.

Patcher mode

Almost broken, it does nothing interesting except, possibly, replace the thread's ServiceTable to an unchanged, clear copy taken from the on-disk image of "ntoskrnl.exe".

Sniffer only IOCTLs:
44038004 - add filtering rule
44038008 - clear stored packet in specified filtering rules list
4403800C - enable specified filtering rule
44038010 - disable specified filtering rule
44038014 - get stored packet from specified filtering rules list
44038018 - process packet like the one received from the wire (filter and store)
4403801C - set maximum rules list length
44038020 - get maximum rules list length
80000004 - enablePacketsFiltering
80000008 - disablePacketsFiltering (PauseSniffer)
800024B4 - send packet to the specified network interface

Common IOCTLs:
80000028 - do nothing (broken/unused part)
80000038 - set external object (broken/unused part)
8000003C - get 4 dwords struct (broken/unused part)
80000040 - copy 260 bytes from the request (broken/unused part)
80000320 - set I/O port mapping (broken/unused part)
80000324 - clear I/O port mapping (broken/unused part)
80000328 - set external PnP Event (broken/unused part)
80000640 - replace specified thread's SDT (ETHREAD.ServiceTable field) to a given copy

Backdoor driven by network sniffer - "mstcp32.sys", "fat32.sys" MD5s 74DE13B5EA68B3DA24ADDC009F84BAEE
B2C7339E87C932C491E34CDCD99FEB07
311D4923909E07D5C703235D83BF4479
21C278C88D8F6FAEA64250DF3BFFD7C6 Size 57 328 - 57 760 bytes Format PE32 Native Compiled 2007.10.02 12:42:14 (GMT)
2001.08.17 20:52:04 (GMT) Version Info
  • FileDescription: TCP/IP driver
  • LegalCopyright: Copyright (C) Microsoft Corp. 1981-1999
  • InternalName: mstcp32.sys

This is a sniffer tool similar to "tdip.sys" and it uses NT4 NDIS-4, XP NDIS-5 interfaces.  It targets incoming traffic on Ethernet and VPN (ndiswanip) interfaces, but instead of dumb packet dumping, it uses received packets as commands for the "process injector" subsystem that is able to extract and execute code from the specially crafted network packets.

Default filtering rules are stored in the "Options" registry value of the driver's registry key. It captures only directed packets (containing a destination address equal to the station address of the NIC).

The driver's filename and device name differ across the samples. They depend on the name of the registry key that is used to start the driver.

Code Patcher

The driver patches OS code to dynamically disable or enable Windows audit logging.

It patches the function "LsapAdtWriteLog" in "lsasrv.dll" module of the "lsass.exe" process.

It searches for pre-defined signatures of the function "LsapAdtWriteLog" of known Windows versions - 4.0, 5.0, 5.1, 5.2 (NT4, Win2000, XP, WinSrv2003).

Then it selects a corresponding offset to replace the opcodes:

  • 'jz' to never taken 'jo' in case of XP
  • jmp over inner logic to procedure epilog in case of Windows Server 2003 so LsapAdtWriteLog skips logging of audit records

The module also patches "SepAdtLogAuditRecord" inside "ntoskrnl.exe" to "retn 4" instead of the first opcode of the function.

The disabled audit can be restored after a timeout or on-event by a dedicated thread.

Expected IOCTL codes:

  • 80000004 - setFilteringRules
  • 80000008 - disablePacketsFiltering (PauseSniffer)
  • 80000028 - do nothing (possible broken GetDriverName)
  • 80000038 - disable_audit
  • 8000003C - enable_audit
Code Injector

The code-builder within this module facilitates exploitation by providing up to four predefined execution templates, which seem to be suitable for generating several code patterns.

Below is a list of the execution templates we found:

  • locate a DLL via PEB structure and resolve exports
  • call single function
  • call four functions
  • call six functions

Using these as a base for the templates, the code-builder inserts parameters and proper offsets to call one of the following code patterns:

  • Locate and call WinExec
  • Locate and call LoadLibraryW, GetProcAddress, call exported procedure, FreeLibrary
  • Locate and call LoadLibraryW, GetProcAddress, call GetModuleHandle, FreeLibrary
  • Locate and call OpenProcess, VirtualAllocEx, WriteProcessMemory, CreateRemoteThread, VirtualFreeEx, CloseHandle

The code injection procedure allocates memory via ZwAllocateVirtualMemory in services.exe and copies implanted code. After that it uses KeInsertQueueApc to let the code run and waits 30 seconds for APC to complete.

When the module starts, it reads registry value [HKLM\System\CurrentControlSet\Services\%driver name%] Processes. This value may contain a list of process names that should be started by injected executable code but only after services.exe and winlogon.exe has been started. The injection of code into winlogon.exe and services.exe ensures that the newly started process will have SYSTEM user privileges. During the injection stage Windows Audit Logging is fully disabled to avoid leaving any suspicious records in Windows Logs.

Magic Packet Recognition

All incoming packets are first filtered by BPF-like rules. The filtering rules are located in [HKLM\System\CurrentControlSet\Services\%driver name%] Options registry value or passed via corresponding IOCTL. Packets that passed through the filter are added in the end of processing queue.

Packets from the queue must have valid checksum values. After checking that, the code XOR-decrypts additional data from the end of the packet. The decrypted end of the packet contains another control structure that defines which encryption algorithm is used to decipher packet body. Supported algorithms include RC5 and RSA. There is a 1024-bits RSA public key hardcoded inside the module body, while a 96-bits RC5 key is generated dynamically.

The backdoor command may arrive in a single packet or be split into pieces and come with several packets. There is a procedure for re-assembling pieces together: a multi-packet command is added to a special packet collector which puts all the pieces together before passing it further.

Backdoor command is stored in the first byte of the decrypted request and can be one of the following values:

Command Byte Short Name Description 0 StatusCheck Reply with encrypted DriverName, packet collector keys, "Processes" and "Params" registry values, names of processes for code injection (defaults are "services.exe" and "winlogon.exe"), list of unprocessed commands from packet collector. 1 Panic Disable packet filtering, securely delete driver file, clear related registry keys, set ClearPageFileAtShutdown flag, unbind adapters, delete devices and prepare for unloading. However, the driver cannot unload itself and it remains in the memory until reboot, which means it can be detected via DRIVER_OBJECT structure. 2 CodeInject Injects code implant received from packet body. 3 CallWinExec Disable Windows Audit and start new process via standard code injection in Winlogon.exe. 4 ResetCollector Drop all packets with specific key from packet collector. 5 Unused or disabled 6 GenerateRC5Key Generates RC5 session key. 96-bits RC5 key is generated by taking a SHA1 hash of value from KeTickCount, XOR-encrypt with SHA1 hash of two PerformanceFrequency values, which is followed by nine more additional SHA1 hashing. 7 SetRegProcesses Sets new "Processes" registry value. 8 DeleteRegProcesses Deletes current "Processes" registry value. 9 SetRegParams Sets a universal "Params" registry value which is used for storing any information, i.e. could be a unique victim marker or exploitation stage number. a Unused or disabled b Ping Reply with classic ICMP Echo Response packet. c ChangeInjectTarget Set new target injection processes (defaults are "Winlogon.exe" and "Services.exe"). Effective until reboot.

Note: "mstcp32" is mentioned together with rootkit-like behavior in 2004 here: http://www.pcreview.co.uk/forums/mstcp32-t1445152.html

Network Sniffer - tdip.sys MD5s 20506375665a6a62f7d9dd22d1cc9870
60dab5bb319281747c5863b44c5ac60d Size 22448 - 28800 bytes Format PE32 Native Compiled 2006.10.16 18:42:40 (GMT)
2003.08.17 21:47:33 (GMT)

Supports the following versions of Windows: NT4 using NDIS-4 and XP using NDIS-5. Doesn't use Vista and later NDIS-6 features. However, later NDIS versions are backward-compatible, so the driver is still valid for current versions of Windows.

Version Info:

  • FileDescription: IP Transport Driver
  • LegalCopyright: © Microsoft Corporation. All rights reserved.
  • FileVersion: 5.1.2600.2180
  • InternalName: tdip.sys

This driver is a packet sniffer for incoming-only traffic on Ethernet and VPN (ndiswanip) interfaces or any used with ms_pschedmp as an alternative connection.

It implements a BPF (Berkeley packet filter) style packet-filtering system that is configured from the driver's registry configuration values or from DeviceIoControl messages.

The captured network packets may be written to disk in libpcap format (magic 0xA1B2C3D4 version 2.4) and encrypted with one-byte XOR, key 0xE3.

The driver's configuration is stored in the registry key:
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\tdip]

  • Options - packet filtering rules in BPF format
  • Tag - selector of filtered packet types / Defaults in case of MediumWan to NDIS_PACKET_TYPE_BROADCAST|NDIS_PACKET_TYPE_MULTICAST|NDIS_PACKET_TYPE_DIRECTED;
    (or NDIS_PACKET_TYPE_BROADCAST|NDIS_PACKET_TYPE_DIRECTED in any other case)
  • ImageFile - full path name to the resulting pcap file
  • Duration - used as Length of the original packet in dump file. (default 0xffff)
  • Backup - max size of the pcap file

IOCTLs:

  • 0x80002004 getCurrentState
  • 0x80002008 setFilteringRules
  • 0x8000200C getFilteringRules
  • 0x80002024 getDumpFileSize
  • 0x80002010/0x80002014/0x80002018/0x8000201C pause/resume
  • 0x80002020 getVersion - returns 2.4.0

Driver has three logical parts, and uses an incomplete function pointer table as interface:

  1. Business logic: filtering rules, packet dumping, device ioctl, options
  2. Ndis driver skeleton
  3. Primitives lib: Strings, XORing, registry I/O

The code is of very good quality. It looks more complicated than Winpcap 2.3 (released 28 mar 2002), but less so than Winpcap 3.0 (released by 10 apr 2003). Interestingly, the driver identifies itself as "version 2.4" in the pcap file despite there being no Winpcap version 2.4.

Key/clipboard logger driver - msrtvd.sys MD5s 98dea1bce37bf7087360e1958400589b
bb8f56874189d5dfe9294f0553a49b83
f6bf3ed3bcd466e5fd1cbaf6ba658716 Size 31 488 - 36 736 bytes Format PE32 Native Compiled 2010.02.19 22:45:18 (GMT)
2008.09.17 16:23:54 (GMT) Version Info
  • FileDescription: MSRTvd interface driver
  • LegalCopyright: © Microsoft Corporation. All rights reserved.
  • InternalName: msrtvd.sys

This is a keylogger and clipboard monitoring tool.

On startup, the driver creates a device named "\Device\Gk0" and a symbolic link named "\DosDevices\Gk".

Then it attaches to the csrss.exe process and disassembles user32.dll and ntdll.dll routines to obtain win32k.sys and ntoskrnl.exe SDT services indexes and pointers of needed Nt/Zw APIs.

Then, using a built-in disassembler, it obtains pointers to NtUserPeekMessage, NtUserGetMessage, NtUserGetClipboardData and using the disassembler again selects the parts of the code that will be then hooked by splicing.

The interceptor routines are copied from a special PE section named ".msda". These routines are able to collect key press chains and clipboard text data, add information about current Time, ProcessName, ForegroundWindowText,and UserName related to this event.

A dedicated thread ("dumper") gathers the collected data, compresses the results with LZO appends it every 30 minutes to a file "%system-wide TEMP%\tm154o.da".

Most strings inside are encrypted by XOR with a pre-seeded random number generator.

IOCTLs:

  • 0x22002C -start dumper thread
  • 0x220030 - stop dumper thread
  • 0x220034 - check if the driver has new data to dump
  • 0x220038 - set two external events signaled on dump data availability (it references a plugin possibility)
  • 0x22003C - restart dumper thread
  • 0x220040 - get size of available data
Collector plugin for Volrec - msrstd.sys MD5s 69e7943f3d48233de4a39a924c59ed2c
15d39578460e878dd89e8911180494ff Size 13 696 - 17 408 bytes Format PE32 Native Compiled 2009.06.05 16:21:55 (GMT)
2009.12.15 16:33:52 (GMT) Version Info
  • FileDescription: msrstd driver
  • LegalCopyright: © Microsoft Corporation. All rights reserved.
  • InternalName: msrstd.sys

This driver is a plugin that collects events from the "volrec.sys" driver, and delivers them by sending DeviceIoControl messages. It collects events about file and disk volume operations.

On startup the driver obtains a pointer to "\Device\volrec", then creates a control device "\Device\msrstd0" and a symbolic link to it named "\DosDevices\msrstd"

All strings inside the driver are encrypted by XOR with a pre-seeded random number generator.

For file events the driver collects the filenames, and caches data about read and write operations. For disk volume events it queries disk properties and reads volume labels and disk serial numbers of removable drives (USB, FireWire drives).

IOCTLs:

0x220004 - turn on VolumeEvents collection
0x220008 - turn off VolumeEvents collection
0x22000C - retrieve previously stored VolumeEvent (operationType, deviceTypeFlags, VolumeLabel, volumeSerialNumber, DosDriveLetter)
0x220010 - turn on FileEvents collection
0x220014 - turn off FileEvents collection
0x220018 - retrieve previously stored FileEvent (fileName, deviceTypeFlags, VolumeLabel, volumeSerialNumber, DosDriveLetter)
0x22001C - connect to Volrec.sys (send ioctl 0x220004), enable plugin operation
0x220020 - disconnect from Volrec.sys (send ioctl 0x220008), disable plugin operation

Filesystem filter driver – volrec.sys, scsi2mgr.sys MD5s a6662b8ebca61ca09ce89e1e4f43665d
c17e16a54916d3838f63d208ebab9879 Size 14 464-14 848 byres Format PE32 Native Compiled 2009.06.05 16:21:57 (GMT)
2009.12.15 16:33:57 (GMT) Version Info
  • FileDescription: Volume recognizer driver
  • LegalCopyright: © Microsoft Corporation. All rights reserved.
  • InternalName: volrec.sys

This driver is a generic filesystem filter which feeds system events to user-mode plugins.

On startup the driver creates a control device named "\Device\volrec" and a symbolic link to it named "\DosDevices\volrec0". It then attaches all available filesystem devices.  It is also, able to handle removable storage devices.

All strings inside the driver are encrypted by XOR with a pre-seeded random number generator.

IOCTLs:

  • 0x220004 - setup plugin interface
  • 0x220008 - disable plugin calls

The driver handles the following system events:

  • file opened, created or closed
  • data is read or written to a file
  • new volume is mounted, unmounted
  • new USB or FireWire device attached
HDD/SSD operation helper driver - WIN32M.SYS MD5s 2b444ac5209a8b4140dd6b747a996653
b3487fdd1efd2d1ea1550fef5b749037 Size 19 456 - 26 631 bytes Format PE32 Native, PE32+ Native Compiled 2001.08.23 17:03:19 (GMT)
2013.05.14 15:58:36 (GMT) Description This module will be the subject of a dedicated blogpost. HDD/SSD firmware operation - nls_933w.dll MD5s 11fb08b9126cdb4668b3f5135cf7a6c5
9f3f6f46c67d3fad2479963361cf118b Size 212 480 - 310 272 bytes Format PE32 DLL, PE32+ DLL Compiled 2010.06.15 16:23:37 (GMT)
2013.05.14 16:12:35 (GMT) Version Info (64bit dll only)
  • FileDescription: Windows Networking Library
  • LegalCopyright: Copyright (C) Microsoft Corp. 1981-2001
  • FileVersion: 80AA
  • InternalName: nls_933w.dll
  • OriginalFilename: nls_933w.dll
  • PrivateBuild: 4.0.1.0
  • ProductName: Microsoft(R) Windows (R) 2000 Operating System
  • ProductVersion: 5.0.2074.0
  • Full Version: 1.0.0.1
Description This (80AA) plugin is a HDD firmware flashing tool which includes an API and the ability to read/write arbitrary information into hidden sectors on the disk.
The plugin will be the subject of a separate blogpost.

Patch Tuesday March 2015 - Stuxnet LNK 0day Fixed

Tue, 03/10/2015 - 20:05

Wait, what? Wasn't the Stuxnet LNK vulnerability CVE-2010-2568, in part reported by Sergey I. Ulasen, patched years ago? Didn't Kim Zetter have enough time to write 448 pages of thoroughly footnoted research on this digital weaponry?

Yes, it was, but MS10-046 didn't completely fix all of the vulnerable code path. And, we just might start to call it the Fanny LNK 0day, after the poorly QA'd USB worm spread across Pakistan using the same LNK exploit. However, we have not observed a newer implementation of this LNK exploit in-the-wild. Yet.

So, machines have remained vulnerable to an actively exploited codebase providing USB support since at least 2008. German researcher Michael Heerklotz reported the remaining flaws in January, and an excellent technical writeup describing his findings is posted on the ZDI blog here. Essentially, an attacker has to create a malicious LNK file with a link path of exactly 257 characters containing embedded unescaped spaces, and two "target" files - one with embedded unescaped spaces and one without. This is not difficult on a usb stick, and it bypasses much of the effective defenses Microsoft has developed for years. "Microsoft has gone to a great deal of effort to make exploitation of memory corruption bugs more difficult. This is a classic example of the Defender's Dilemma -- the defender must be strong everywhere, while the attacker needs to find only one mistake." In this case, it's more that the attacker had to chain together a complex series of overlooked steps.

Microsoft's release of thirteen other bulletins includes a large rollup of fixes for RCE across all versions of Internet Explorer, IE6 - IE11. This MS15-018 bulletin is rated critical, and it requires a reboot.

Pages