Malware RSS Feed
In the fall of 2014, we discovered a new banking Trojan, which caught our attention for two reasons:
- First, it is interesting from the technical viewpoint, because it uses a new technique for loading modules.
- Second, an analysis of its configuration files has shown that the malware targets a large number of online-banking systems: over 150 different banks and 20 payment systems in 15 countries. Banks in the UK, Spain, the US, Russia, Japan and Italy make up the majority of its potential targets.
Kaspersky Lab products detect the new banking malware as Trojan-Banker.Win32.Chthonic.
The Trojan is apparently an evolution of ZeusVM, although it has undergone a number of significant changes. Chthonic uses the same encryptor as Andromeda bots, the same encryption scheme as Zeus AES and Zeus V2 Trojans, and a virtual machine similar to that used in ZeusVM and KINS malware.Infection
We have seen several techniques used to infect victim machines with Trojan-Banker.Win32.Chthonic:
- sending emails containing exploits;
- downloading the malware to victim machines using the Andromeda bot (Backdoor.Win32.Androm in Kaspersky Lab classification).
When sending messages containing an exploit, cybercriminals attached a specially crafted RTF document, designed to exploit the CVE-2014-1761 vulnerability in Microsoft Office products. The file has a .DOC extension to make it look less suspicious.
Sample message with CVE-2014-1761 exploit
In the event of successful vulnerability exploitation, a downloader for the Trojan was downloaded to the victim computer. In the example above, the file is downloaded from a compromised site – hxxp://valtex-guma.com.ua/docs/tasklost.exe.
The Andromeda bot downloaded the downloader from hxxp://globalblinds.org/BATH/lider.exe.Downloading the Trojan
Once downloaded, the downloader injects its code into the msiexec.exe process. It seems that the downloader is based on the Andromeda bot's source code, although the two use different communication protocols.
Example of common functionality of Andromeda and Chthonic downloaders
Differences in communication protocols used by Andromeda and Chthonic C&C
The Chthonic downloader contains an encrypted configuration file (similar encryption using a virtual machine was used in KINS and ZeusVM). The main data contained in the configuration file includes: a list of С&С servers, a 16-byte key for RC4 encryption, UserAgent, botnet id.
The main procedure of calling virtual machine functions
After decrypting the configuration file, its individual parts are saved in a heap - in the following format:
This is done without passing pointers. The bot finds the necessary values by examining each heap element using the RtlWalkHeap function and matching its initial 4 bytes to the relevant MAGIC VALUE.
The downloader puts together a system data package typical of ZeuS Trojans (local_ip, bot_id, botnet_id, os_info, lang_info, bot_uptime and some others) and encrypts it first using XorWithNextByte and then using RC4. Next, the package is sent to one of the C&C addresses specified in the configuration file.
In response, the malware receives an extended loader – a module in a format typical of ZeuS, i.e., not a standard PE file but a set of sections that are mapped to memory by the loader itself: executable code, relocation table, point of entry, exported functions, import table.
Code with section IDs matching the module structures
It should be noted that the imports section includes only API function hashes. The import table is set up using the Stolen Bytes method, using a disassembler included in the loader for this purpose. Earlier, we saw a similar import setup in Andromeda.
Fragment of the import setup function in Andromeda and Chthonic
Header of a structure with module
The extended loader also contains a configuration file encrypted using the virtual machine. It loads the Trojan's main module, which in turn downloads all the other modules. However, the extended loader itself uses AES for encryption, and some sections are packed using UCL. The main module loads additional modules and sets up import tables in very much the same way as the original Chthonic downloader, i.e. this ZeuS variant has absorbed part of the Andromeda functionality.
The entire sequence in which the malware loads, including the modules that are described below, is as follows:
Trojan-Banker.Win32.Chthonic has a modular structure. To date, we have discovered the following modules:Name Description Has a 64bit version main Main module (v188.8.131.52 - v184.108.40.206) Yes info Collects system information Yes pony Module that steals saved passwords No klog Keylogger Yes http Web injection and formgrabber module Yes vnc Remote access Yes socks Proxy server Yes cam_recorder Recording video from the web camera Yes
The impressive set of functions enables the malware to steal online banking credentials using a variety of techniques. In addition, VNC and cam recorder modules enable attackers to connect to the infected computer remotely and use it to carry out transactions, as well as recording video and sound if the computer has a webcam and microphone.Injections
Web injections are Chthonic's main weapon: they enable the Trojan to insert its own code and images into the code of pages loaded by the browser. This enables the attackers to obtain the victim's phone number, one-time passwords and PINs, in addition to the login and password entered by the victim.
For example, for one of the Japanese banks the Trojan hides the bank's warnings and injects a script that enables the attackers to carry out various transactions using the victim's account:
Online banking page screenshots before and after the injection
Interesting functions in injected script
The script can also display various fake windows in order to obtain the information needed by the attackers. Below is an example of a window which displays a warning of non-existent identification problems and prompts the user to enter TAN:
Fake TAN entry window
Our analysis of attacks against customers of Russian banks has uncovered an unusual web injection scenario. When opening an online banking web page in the browser, the entire contents of the page is spoofed, not just parts of it as in an ordinary attack. From the technical viewpoint, the Trojan creates an iframe with a phishing copy of the website that has the same size as the original window.
Below is a fragment of injected code, which replaces everything between title and body closing tags with the following text:
And here is the script itself:
Additionally, the bot receives a command to establish a backconnect connection if the injection is successful:
There are several botnets with different configuration files. Overall, the botnets we are aware of target online banking systems of over 150 different banks and 20 payment systems in 15 countries. The cybercriminals seem most interested in banks in the UK, Spain, the US, Russia, Japan and Italy.
Chtonic target distribution by country
It is worth noting that, in spite of the large number of targets on the list, many code fragments used by the Trojan to perform web injections can no longer be used, because banks have changed the structure of their pages and, in some cases, the domains as well. It should also be noted that we saw some of these fragments in other bots' config files (e.g., Zeus V2) a few years back.Conclusion
We can see that the ZeuS Trojan is still actively evolving and its new implementations take advantage of cutting-edge techniques developed by malware writers. This is significantly helped by the ZeuS source code having been leaked. As a result, it has become a kind of framework for malware writers, which can be used by anyone and can easily be adapted to cybercriminals' new needs. The new Trojan – Chthonic – is the next stage in the evolution of ZeuS: it uses Zeus AES encryption, a virtual machine similar to that used by ZeusVM and KINS, and the Andromeda downloader.
What all of this means is that we will undoubtedly see new variants of ZeuS in the future.A few md5:
Over the past years, Kaspersky's Global Research and Analysis Team (GReAT) has shed light on some of the biggest APT campaigns, including RedOctober, Flame, NetTraveler, Miniduke, Epic Turla, Careto/Mask and others. While studying these campaigns we have also identified a number of 0-day exploits, including the most recent CVE-2014-0546. We were also among the first to report on emerging trends in the APT world, such as cyber mercenaries who can be contracted to launch lightning attacks or more recently, attacks through unusual vectors such as hotel Wi-Fi. Over the past years, Kaspersky Lab's GReAT team has monitoring more than 60 threat actors responsible for cyber-attacks worldwide, organizations which appear to be fluent in many languages such as Russian, Chinese, German, Spanish, Arabic, Persian and others.
By closely observing these threat actors, we put together a list of what appear to be the emerging threats in the APT world. We think these will play an important role in 2015 and deserve special attention, both from an intelligence point of view but also with technologies designed to stop them.The merger of cyber-crime and APT
For many years, cyber-criminal gangs focused exclusively on stealing money from end users. An explosion of credit card theft, hijacking of electronic payment accounts or online banking connections led to consumer losses in the worth hundreds of millions of dollars. Maybe this market is no longer so lucrative, or maybe the cybercriminal market is simply overcrowded, but it now seems like there is a struggle being waged for 'survival'. And, as usual, that struggle is leading to evolution.
What to expect: In one incident we recently investigated attackers compromised an accountant's computer and used it to initiate a large transfer with their bank. Although it might seem that this is nothing very unusual, we see a more interesting trend: Targeted attacks directly against banks, not their users.
In a number of incidents investigated by Kaspersky Lab experts from the Global Research and Analysis Team, several banks were breached using methods straight out of the APT playbook. Once the attackers got into the banks' networks, they collected enough information to enable them to steal money directly from the bank in several ways:
- Remotely commanding ATMs to dispense cash.
- Performing SWIFT transfers from various customer accounts,
- Manipulating online banking systems to perform transfers in the background.
These attacks are an indication of a new trend that is embracing APT style attacks in the cybercriminal world. As usual, cybercriminals prefer to keep it simple: they now attack the banks directly because that's where they money is. We believe this is a noteworthy trend that will become more prominent in 2015.Fragmentation of bigger APT groups
2014 saw various sources expose APT groups to the public eye. Perhaps the best-known case is the FBI indictment of five hackers on various computer crimes:
This public "naming and shaming" means we expect some of the bigger and "noisier" APT groups to shatter and break into smaller units, operating independently.
What to expect: This will result in a more widespread attack base, meaning more companies will be hit, as smaller groups diversify their attacks. At the same time, it means that bigger companies that were previously compromised by two or three major APT groups (eg. Comments Crew and Wekby) will see more varied attacks from a wider range of sources.Evolving malware techniques
As computers become more sophisticated and powerful, operating systems also become more complex. Both Apple and Microsoft have spent a lot of time improving the security posture of their respective operating systems. Additionally, special tools such as Microsoft's EMET are now available to help thwart targeted attacks against software vulnerabilities.
With Windows x64 and Apple Yosemite becoming more popular, we expect APT groups to update their toolsets with more powerful backdoors and technologies to evade security solutions.
What to expect: Today, we are already seeing APT groups constantly deploying malware for 64-bit systems, including 64-bit rookits. In 2015, we expect to see more sophisticated malware implants, enhanced evasion techniques and more use of virtual file systems (such as those from Turla and Regin) to conceal precious tools and stolen data.
While we see these increases in advanced techniques, some attackers are moving in the opposite direction. While minimizing the number of exploits and amount of compiled code they introduce to compromised networks altogether, their work continues to require sophisticated code or exploit introduction at a stable entry into the enterprise, script tools and escalation of privilege of all sorts, and stolen access credentials at victim organizations.
As we saw with BlackEnergy 2 (BE2), attackers will actively defend their own presence and identity within victim networks once discovered. Their persistence techniques are becoming more advanced and expansive. These same groups will step up the amount and aggression of destructive last effort components used to cover their tracks, and they include more *nix support, networking equipment, and embedded OS support. We have already seen some expansion from BE2, Yeti, and Winnti actors.New methods of data exfiltration
The days when attackers would simply activate a backdoor in a corporate network and start siphoning terabytes of information to FTP servers around the world are long gone. Today, more sophisticated groups use SSL on a regular basis alongside custom communication protocols.
Some of the more advanced groups rely on backdooring networking devices and intercepting traffic directly for commands. Other techniques we have seen include exfiltration of data to cloud services, for instance via the WebDAV protocol (facilitates collaboration between users in editing and managing documents and files stored on web servers).
These in turn have resulted in many corporations banning public cloud services such as Dropbox from their networks. However, this remains an effective method of bypassing intrusion detection systems and DNS blacklists.
What to expect: In 2015, more groups to adopt use of cloud services in order to make exfiltration stealthier and harder to notice.New APTs from unusual places as more countries join the cyber arms race
In February 2014, we published research into Careto/Mask, an extremely sophisticated threat actor that appears to be fluent in Spanish, a language rarely seen in targeted attacks. In August, we also released a report on Machete, another threat actor using the Spanish language.
Before that, we were accustomed to observing APT actors and operators that are fluent in relatively few languages. Additionally, many professionals do not use their native language, preferring instead to write in perfect English.
In 2014, we observed a lot of nations around the world publicly expressing an interest in developing APT capabilities:
What to expect: Although we haven't yet seen APT attacks in Swedish, we do predict that more nations will join the "cyber-arms" race and develop cyber-espionage capabilities.Use of false flags in attacks
Attackers make mistakes. In the vast majority of the cases we analyze, we observe artifacts that provide clues about the language spoken by the attackers. For instance, in the case of RedOctober and Epic Turla, we concluded that the attackers were probably fluent in the Russian language. In the case of NetTraveler we came to the conclusion that attackers were fluent in Chinese.
In some cases, experts observe other meta features that could point toward the attackers. For example, performing file timestamp analysis of the files used in an attack may lead to the conclusion in what part of the world most of the samples were compiled.
However attackers are beginning to react to this situation. In 2014 we observed several "false flag" operations where attackers delivered "inactive" malware commonly used by other APT groups. Imagine a threat actor of Western origin dropping a malware commonly used by a "Comment Crew," a known Chinese threat actor. While everyone is familiar with the "Comment Crew" malware implants, few victims could analyze sophisticated new implants. That can easily mislead people into concluding that the victim was hit by the Chinese threat actor.
What to expect: In 2015, with governments increasingly keen to "name and shame" attackers, we believe that APT groups will also carefully adjust their operations and throw false flags into the game.Threat actors add mobile attacks to their arsenal
Although APT groups have been observed infecting mobile phones, this hasn't yet become a major trend. Perhaps the attackers wish to get data that isn't usually available on mobiles, or maybe not all of them have access to the technologies that can infect Android and iOS devices.
In 2014 we saw several new APT tools designed for infecting mobiles, for instance Hacking Team's Remote Control System mobile modules.
Additionally, during the Hong Kong protests in October 2014, attacks were seen against Android and iOS users which appear to be connected to APT operations.
Although a mobile phone might not have valuable documents and schematics, or geopolitical expansion plans for next 10 years, they can be a valuable source of contacts as well as listening points. We observed this with the RedOctober group, which had the ability to infect mobile phones and turn them into "Zakladka's", mobile bugs.
What to expect: In 2015, we anticipate more mobile-specific malware, with a focus on Android and jailbroken iOS.APT+Botnet: precise attack + mass surveillance
In general, APT groups are careful to avoid making too much noise with their operations. This is why the malware used in APT attacks is much less widespread than common crimeware such as Zeus, SpyEye and Cryptolocker.
In 2014 we observed two APT groups (Animal Farm and Darkhotel) using botnets in addition to their regular targeted operations. Of course, botnets can prove to be a vital asset in cyberwar and can be used to DDoS hostile countries; this has happened in the past. We can therefore understand why some APT groups might want to build botnets in addition to their targeted operations.
In addition to DDoS operations, botnets can also offer another advantage - mass surveillance apparatus for a "poor country". For instance, Flame and Gauss, which we discovered in 2012, were designed to work as a mass surveillance tool, automatically collecting information from tens of thousands of victims. The information would have to be analyzed by a supercomputer, indexed and clustered by keywords and topics; most of it would probably be useless. However, among those hundreds of thousands of exfiltrated documents, perhaps one provides key intelligence details, that could make a difference in tricky situations.
What to expect: In 2015 more APT groups will embrace this trend of using precise attacks along with noisy operations and deploy their own botnets.Targeting of hotel networks
The Darkhotel group is one of the APT actors known to have targeted specific visitors during their stay in hotels in some countries. Actually, hotels provide an excellent way of targeting particular categories of people, such as company executives. Targeting hotels is also highly lucrative because it provides intelligence about the movements of high profile individuals around the world.
Compromising a hotel reservation system is an easy way to conduct reconnaissance on a particular target. It also allows the attackers to know the room where the victim is staying, opening up the possibility of physical attacks as well as cyber-attacks.
It isn't always easy to target a hotel. This is why very few groups, the elite APT operators, have done it in the past and will use it as part of their toolset.
What to expect: In 2015, a few other groups might also embrace these techniques, but it will remain beyond the reach of the vast majority of APT players.Commercialization of APT and the private sector
Over the last few years, we published extensive research into malware created by companies such as HackingTeam or Gamma International, two of the best known vendors of "legal spyware". Although these companies claim to sell their software only to "trusted government entities", public reports from various sources, including Citizen Lab, have repeatedly shown that spyware sales cannot be controlled. Eventually, these dangerous software products end up in the hands of less trustworthy individuals or nations, who can use them for cyber-espionage against other countries or their own people.
The fact is that such activities are highly profitable for the companies developing the cyber-espionage software. They are also low risk because – so far – we have not seen a single case where one of these companies was convicted in a cyber-espionage case. The developers of these tools are usually out of the reach of the law, because the responsibility falls with the tool users, not the company that develops and facilitates the spying.
What to expect: It's a high-reward, low risk business that will lead to the creation of more software companies entering the "legal surveillance tools" market. In turn, these tools will be used for nation-on-nation cyber-espionage operations, domestic surveillance and maybe even sabotage.Conclusions
In general, 2014 was a rather sophisticated and diverse year for APT incidents. We discovered several zero-days, for instance CVE-2014-0515 which was used by a group we call "Animal Farm". Another zero-day we discovered was CVE-2014-0487, used by the group known as DarkHotel. In addition to these zero-days, we observed several new persistence and stealth techniques, which in turn resulted in the development and deployment of several new defense mechanisms for our users.
If we can call 2014 "sophisticated", the word for 2015 will be "elusive". We believe that more APT groups will become concerned with exposure and they will take more advanced measures to hide from discovery.
Finally, some of them will deploy false flag operations. We anticipate these developments and, as usual, will document them thoroughly in our reports.
Two years ago, we published our research into RedOctober, a complex cyber-espionage operation targeting diplomatic embassies worldwide. We named it RedOctober because we started this investigation in October 2012, an unusually hot month.
After our announcement in January 2013, the RedOctober operation was promptly shut down and the network of C&Cs was dismantled. As usually happens which these big operations, considering the huge investment and number of resources behind it, they don't just "go away" forever. Normally, the group goes underground for a few months, redesigns the tools and the malware and resume operations.
Since January 2013, we've been on the lookout for a possible RedOctober comeback. One possible hit was triggered when we observed Mevade, an unusual piece of malware that appeared late in 2013. The Mevade C&C name styles as well as some other technical similarities indicated a connection to RedOctober, but the link was weak. It wasn't until August 2014 that we observed something which made us wonder if RedOctober is back for good.Meet Cloud Atlas
In August 2014, some of our users observed targeted attacks with a variation of CVE-2012-0158 and an unusual set of malware. We did a quick analysis of the malware and it immediately stood out because of certain unusual things that are not very common in the APT world.
Some of the filenames used in the attacks included:
- FT - Ukraine Russia's new art of war.doc
- Катастрофа малайзийского лайнера.doc
- Diplomatic Car for Sale.doc
- Organigrama Gobierno Rusia.doc
- Информационное письмо.doc
- Форма заявки (25-26.09.14).doc
- Информационное письмо.doc
- Car for sale.doc
- Af-Pak and Central Asia's security issues.doc
At least one of them immediately reminded us of RedOctober, which used a very similarly named spearphish: "Diplomatic Car for Sale.doc". As we started digging into the operation, more details emerged which supported this theory.
Perhaps the most unusual fact was that the Microsoft Office exploit didn't directly write a Windows PE backdoor on disk. Instead, it writes an encrypted Visual Basic Script and runs it.
Cloud Atlas exploit payload - VBScript
This VBScript drops a pair of files on disk - a loader and an encrypted payload. The loader appears to be different every time and internal strings indicate it is "polymorphically" generated. The payload is always encrypted with a unique key, making it impossible to decrypt unless the DLL is available.
We observed several different spear-phishing documents that drop uniquely named payloads. For instance, the "qPd0aKJu.vbs" file MD5:
E211C2BAD9A83A6A4247EC3959E2A730 drops the following files:
DECF56296C50BD3AE10A49747573A346 - bicorporate - encrypted payload
D171DB37EF28F42740644F4028BCF727 - ctfmonrn.dll - loader
The VBS also adds a registry key:
HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Run\ setting the key "bookstore" to the value "regsvr32 %path%\ctfmonrn.dll /s", which ensures the malware runs every time at system boot.
Some of the DLL names we observed include:f4e15c1c2c95c651423dbb4cbe6c8fd5 - bicorporate.dll
649ff144aea6796679f8f9a1e9f51479 - fundamentive.dll
40e70f7f5d9cb1a669f8d8f306113485 - papersaving.dll
58db8f33a9cdd321d9525d1e68c06456 - previliges.dll
f5476728deb53fe2fa98e6a33577a9da - steinheimman.dll
Some of the payload names include:steinheimman
The payload includes an encrypted configuration block which contains information about the C&C sever:
The information from the config includes a WebDAV URL which is used for connections, a username and password, two folders on the WebDAV server used to store plugins/modules for the malware and where data from the victim should be uploaded.C&C communication
The Cloud Atlas implants utilize a rather unusual C&C mechanism. All the malware samples we've seen communicate via HTTPS and WebDav with the same server "cloudme.com", a cloud services provider. According to their website, CloudMe is owned and operated by CloudMe AB, a company based in Linköping, Sweden.
(Important note: we do not believe that CloudMe is in any way related to the Cloud Atlas group - the attackers simply create free accounts on this provider and abuse them for command-and-control).
Each malware set we have observed so far communicates with a different CloudMe account though. The attackers upload data to the account, which is downloaded by the implant, decrypted and interpreted. In turn, the malware uploads the replies back to the server via the same mechanism. Of course, it should be possible to reconfigure the malware to use any Cloud-based storage service that supports WebDAV.
Here's a look at one such account from CloudMe:
The data from the account:
The files stored in the randomly named folder were uploaded by the malware and contain various things, such as system information, running processes and current username. The data is compressed with LZMA and encrypted with AES, however, the keys are stored in the malware body which makes it possible to decrypt the information from the C&C.
We previously observed only one other group using a similar method – ItaDuke – that connected to accounts on the cloud provider mydrive.ch.Victim statistics: top 5 infected countries CloudAtlas RedOctober Russia 15 35 Kazakhstan 14 21 Belarus 4 5 India 2 14 Czech Republic 2 5 Similarities with RedOctober
Just like with RedOctober, the top target of Cloud Atlas is Russia, followed closely by Kazakhstan, according to data from the Kaspersky Security Network (KSN). Actually, we see an obvious overlap of targets between the two, with subtle differences which closely account for the geopolitical changes in the region that happened during the last two years.
Interestingly, some of the spear-phishing documents between Cloud Atlas and RedOctober seem to exploit the same theme and were used to target the same entity at different times.Cloud Atlas RedOctober
Both Cloud Atlas and RedOctober malware implants rely on a similar construct, with a loader and the final payload that is stored encrypted and compressed in an external file. There are some important differences though, especially in the encryption algorithms used – RC4 in RedOctober vs AES in Cloud Atlas.
The usage of the compression algorithms in Cloud Altas and RedOctober is another interesting similarity. Both malicious programs share the code for LZMA compression algorithm. In CloudAtlas it is used to compress the logs and to decompress the decrypted payload from the C&C servers, while in Red October the "scheduler" plugin uses it to decompress executable payloads from the C&C.
It turns out that the implementation of the algorithm is identical in both malicious modules, however the way it is invoked is a bit different, with additional input sanity checks added to the CloudAtlas version.
Another interesting similarity between the malware families is the configuration of the build system used to compile the binaries. Every binary created using the Microsoft Visual Studio toolchain has a special header that contains information about the number of input object files and version information of the compilers used to create them, the "Rich" header called so by the magic string that is used to identify it in the file.
We have been able to identify several RedOctober binaries that have "Rich" headers describing exactly the same layout of VC 2010 + VC 2008 object files. Although this doesn't necessarily mean that the binaries were created on the same development computer, they were definitely compiled using the same version of the Microsoft Visual Studio up to the build number version and using similar project configuration.Number of object files, CloudAtlas loader Number of object files, Red October Office plugin Number of object files,Red October Fileputexec plugin HEX compiler version Decoded compiler version 01 01 01 009D766F VC 2010 (build 30319) 01 01 01 009B766F VC 2010 (build 30319) 22 2E 60 00AB766F VC 2010 (build 30319) 5B 60 A3 00010000 – 05 07 11 00937809 VC 2008 (build 30729) 72 5C AD 00AA766F VC 2010 (build 30319) 20 10 18 009E766F VC 2010 (build 30319)
To summarize the similarities between the two:Cloud Atlas RedOctober Shellcode marker in spearphished documents PT@T PT@T Top target country Russia Russia Compression algorithm used for C&C communications LZMA LZMA C&C servers claim to be / redirect to BBC (mobile malware) BBC Compiler version VC 2010 (build 30319) VC 2010 (build 30319) (some modules)
Finally, perhaps the strongest connection comes from targeting. Based on observations from KSN, some of the victims of RedOctober are also being targeted by CloudAtlas. In at least one case, the victim's computer was attacked only twice in the last two years, with only two malicious programs – RedOctober and Cloud Atlas.
These and other details make us believe that CloudAtlas represents a rebirth of the RedOctober attacks.Conclusion
Following big announcements and public exposures of targeted attack operations, APT groups behave in a predictable manner. Most Chinese-speaking attackers simply relocate C&C servers to a different place, recompile the malware and carry on as if nothing happened.
Other groups that are more nervous about exposure go in a hibernation mode for months or years. Some may never return using the same tools and techniques.
However, when a major cyber-espionage operation is exposed, the attackers are unlikely to completely shut down everything. They simply go offline for some time, completely reshuffle their tools and return with rejuvenated forces.
We believe this is also the case of RedOctober, which makes a classy return with Cloud Atlas.
Kaspersky products detect the malware from the Cloud Atlas toolset with the following verdicts:
Several days ago, our products detected an unusual sample from the Destover family. The Destover family of trojans has been used in the high profile attacks known as DarkSeoul, in March 2013, and more recently, in the attack against Sony pictures in November 2014. We wrote about it on December 4th, including the possible links with the Shamoon attack from 2012.
The new sample is unusual in the sense it is signed by a valid digital certificate from Sony:
The signed sample has been previously observed in a non signed form, as MD5: 6467c6df4ba4526c7f7a7bc950bd47eb and appears to have been compiled in July 2014.
The new sample has the MD5 e904bf93403c0fb08b9683a9e858c73e and appears to have been signed on December 5th, 2014, just a few days ago.
Functionally, the backdoor contains two C&Cs and will alternately try to connect to both, with delays between connections:
- 208.105.226[.]235:443 - United States Champlain Time Warner Cable Internet Llc
- 203.131.222[.]102:443 - Thailand Bangkok Thammasat University
So what does this mean? The stolen Sony certificates (which were also leaked by the attackers) can be used to sign other malicious samples. In turn, these can be further used in other attacks. Because the Sony digital certificates are trusted by security solutions, this makes attacks more effective. We've seen attackers leverage trusted certificates in the past, as a means of bypassing whitelisting software and default-deny policies.
We've already reported the digital certificate to COMODO and Digicert and we hope it will be blacklisted soon. Kaspersky products will still detect the malware samples even if signed by digital certificates.
Stolen certificate serial number:
- 01 e2 b4 f7 59 81 1c 64 37 9f ca 0b e7 6d 2d ce
- 8d f4 6b 5f da c2 eb 3b 47 57 f9 98 66 c1 99 ff 2b 13 42 7a
The end of the year is traditionally a time for reflection – for taking stock of our lives before considering what lies ahead. We'd like to offer our customary retrospective of the key events that shaped the threat landscape in 2014.1. Targeted attacks and malware campaigns
Targeted attacks are now an established part of the threat landscape, so it's no surprise to see them feature in our yearly review.
The complex cyber-espionage campaign called 'Careto' or 'The Mask' (Careto is Spanish slang for 'ugly face' or 'mask') was designed to steal sensitive data from specific organizations. The victims of the attack included government agencies, embassies, energy companies, research institutions, private equity firms and activists from 31 countries around the world. Careto included a sophisticated backdoor Trojan capable of intercepting all communication channels and of harvesting all kinds of data from infected computers – including encryption keys, VPN configurations, SSH keys, RDP files and some unknown file types that could be related to bespoke military/government-level encryption tools. The code was highly modular, allowing the attackers to add new functionality at will. There are versions of the backdoor for Windows and Mac OS X and we also found references in some modules indicating that there might be versions for Linux, iOS and Android. As with any sophisticated campaign of this sort, attribution is difficult. Use of the Spanish language in the code doesn't help, since Spanish is spoken in many parts of the world. Also, it's possible that its use is an intentional piece of misdirection. However, the very high degree of professionalism of the group behind this attack is unusual for cybercriminal groups – one indicator that Careto could be a state-sponsored campaign. Like previous targeted attack campaigns, the roots of Careto stretch back well before the threat first came to light: we believe that the attackers have been active since 2007.
Early in March there was widespread discussion among security researchers about a cyber-espionage campaign called 'Epic Turla'. Researchers at G-DATA believed the malware may have been created by Russian special services; while research carried out by BAE Systems linked it to malware identified as 'Agent.btz' that dates back to 2007 and was used in 2008 to infect the local networks of US military operations in the Middle East. Our initial analysis of Epic Turla focused on the malware's use of USB flash drives to store stolen data that can't be sent directly over the Internet to the attackers' Command-and-Control (C2) server. The worm writes a file called 'thumb.dd' to all USB flash drives connected to an infected computer. If the flash drive is subsequently inserted into another computer, the 'thumb.dd' file is copied to the new computer. Epic Turla isn't the only malware that is aware of 'thumb.dd'. This is one of the files in the 'USB Stealer module' in Red October. Looking back further, Gauss and miniFlame were aware of 'thumb.dd and looked for the file on USB flash drives. You can find a chart showing the points of comparison here. We think it's likely that there are tens of thousands of USB flash drives around the world containing files called 'thumb.dd' created by this malware.
In our subsequent analysis of Epic Turla we explained how the attackers use social engineering to spread the malware and highlighted the overall structure of the campaign. The attackers use spear-phishing emails to trick their victims into installing a backdoor on their computer. Some of these include zero-day exploits – one affecting Adobe Acrobat Reader and the other a privilege escalation vulnerability in Windows XP and Windows Server 2003. They also use watering-hole attacks that deploy a Java exploit, Adobe Flash exploits and Internet Explorer exploits, or trick victims into running fake 'Flash Player' malware installers. Depending on the IP address of the victim, the attackers serve Java or browser exploits, signed fake Adobe Flash Player software or a fake version of Microsoft Security Essentials. Unsurprisingly, the choice of web sites reflects the specific interests of the attackers (as well as the interests of the victims). However, our analysis showed that the Epic Turla backdoor is just the first stage of the infection. It is used to deploy a more sophisticated backdoor known as the 'Cobra/Carbon system' (named 'Pfinet' by some anti-malware products). The unique knowledge to operate these two backdoors indicates a clear and direct connection between them: one is used to gain a foothold and validate the high-profile victim. If the victim proves to be of interest to the attackers, the compromised computer is upgraded to the full Carbon system. You can find an overview of the Epic Turla campaign here:
In June we reported on our research into an attack on the clients of a large European bank that resulted in the theft of half a million euros in just one week. We named this 'Luuuk', after the path in the administration panel used in the C2 server. Although we were unable to obtain the malware used to infect the victims, we believe the criminals used a banking Trojan that performed 'Man-in-the-Browser' operations to steal the victims' credentials through a malicious web injection. Based on the information available in some of the log files, the malware stole usernames, passwords and one-time passcodes (OTP) in real time. The attackers used the stolen credentials to check the victim's account balance and perform malicious transactions automatically, probably operating in the background of a legitimate banking session. The stolen money was then transferred automatically to pre-defined money mule accounts. The classification of pre-defined money mules used by the attackers was very interesting. There were four different money mule groups, each defined by the amount of money the mules in the group could accept – probably a reflection of the level of trust between them. We identified 190 victims in total, most of them located in Italy and Turkey. The sums stolen from each victim ranged from €1,700 to €39,000; and amounted to €500,000.
Although the attackers removed all sensitive components soon after our investigation started, we believe that this represents a change of infrastructure rather than a complete shutdown of the operation. The cybercriminals behind the campaign are highly professional and very active. They have also shown proactive operational security activities, changing tactics and removing traces when discovered. The investigation into this campaign, which we reported to the bank concerned and to the appropriate law enforcement agencies, is ongoing.
The end of June saw the re-activation of a targeted attack campaign from early 2013, called 'MiniDuke'. The original campaign stood out for several reasons. It included a custom backdoor written in the 'old school' Assembler programming language. The attack was managed using an unusual command-and-control (C2) infrastructure: it made use of multiple redundancy paths, including Twitter accounts. The developers transferred their updated executables hidden inside GIF files.
Targets of the new operation, known as 'CosmicDuke', or 'TinyBaron', include government, diplomatic, energy, military and telecom operators. But unusually the list of victims also includes those involved in the trafficking and reselling of illegal substances, including steroids and hormones. It's not clear why: maybe the customizable backdoor was made available as so-called 'legal spyware', or it was available in the underground market and was purchased by various rivals in the pharmaceutical business to spy on each other.
Victim geography (Miniduke and CosmicDuke)
The malware spoofs popular applications designed to run in the background - including file information, icons and even file size. The backdoor itself is compiled using 'BotGenStudio' - a customizable framework that allows the attackers to enable and disable components when the bot is constructed. The malware not only steals files with specific extensions, but also harvests passwords, history, network information, address books, information displayed on the screen (screenshots are made every five minutes) and other sensitive data. Each victim is assigned a unique ID, making it possible to push specific updates to individual victims.
The malware is protected with a custom obfuscated loader which heavily consumes CPU resources for 3-5 minutes before passing execution to the payload. This makes it hard to analyze. But it also drains the resources needed by security software to emulate the malware's execution. On top of its own obfuscator, the malware makes heavy use of encryption and compression based on the RC4 and LZRW algorithms. They are implemented slightly differently to the standard versions - we believe that this is done deliberately to mislead researchers. The internal configuration of the malware is encrypted, compressed and serialized as a complicated registry-like structure, which has various record types including strings, integers and internal references. Stolen data uploaded to the C2 server is split into small chunks (of around 3KB), which are compressed, encrypted and placed in a container to be uploaded to the server. If it's a large file, it may be placed into several hundred different containers that are all uploaded independently. It's likely that these data chunks are parsed, decrypted, unpacked, extracted and reassembled on the attacker's side. While this method might add an overhead, the layers of additional processing ensure that very few researchers will get to the original data. This method also offers increased reliability against network errors.
In July we published an in-depth analysis of a targeted attack campaign that we dubbed 'Crouching Yeti' – also known as 'Energetic Bear', because researchers from CrowdStrike had suggested that the attackers were located in Russia: we don't think there's enough evidence to confirm this one way or the other. This campaign, active since late 2010, has so far targeted the following sectors: industrial/machinery, manufacturing, pharmaceutical, construction, education and information technology. So far there have been more than 2,800 victims worldwide, and we have been able to identify 101 different victim organizations – mostly in the United States, Spain, Japan, Germany, France, Italy, Turkey, Ireland, Poland and China.
The attackers behind Crouching Yeti use various types of malware (all designed to infect systems running Windows) to infiltrate their victims, extend their reach within the target organizations and steal confidential data, including intellectual property and other strategic information. The malware used includes special modules to collect data from specific industrial IT environments. Infected computers connect to a large network of hacked web sites that host malware modules, hold information about victims and send commands to infected systems. The attackers use three methods to infect their victims. These include a legitimate software installer re-packaged to include a malicious DLL file; spear-phishing e-mails; and watering-hole attacks.
Technology is now an integral part of our lives, so it's hardly surprising to see a cyber-dimension to conflicts around the world. This is especially true of the Middle East, where geo-political conflicts have intensified in recent years. In August we reported on the increase in malware activity in Syria from early 2013. The victims of these attacks are not only located in Syria: the malware has also been seen in Turkey, Saudi Arabia, Lebanon, Palestine, the United Arab Emirates, Israel, Morocco, France and the United States. We were able to track the C2 servers of the attackers to IP addresses in Syria, Russia, Lebanon, the United States and Brazil. In total, we found 110 files, 20 domains and 47 IP addresses associated with the attacks.
It's clear that the groups involved in the attacks are well organized. So far the attackers have made use of established malware tools rather than developing their own (although they use a variety of obfuscation methods to bypass simple signature-based detection). However, we think it's likely that the number and sophistication of malware used in the region is likely to increase.
In November we published our analysis of the 'Darkhotel' APT, a campaign that has been operating for almost a decade, targeting thousands of victims across the globe. 90% of the infections we have seen are in Japan, Taiwan, China, Russia and Hong Kong, but we have also seen infections in Germany, the USA, Indonesia, India, and Ireland.
The campaign employs varying degrees of targeting. First, they use spear-phishing e-mails and zero-day exploits to infiltrate organizations from different sectors, including Defense Industrial Base (DIB), government and Non-Governmental Organizations (NGOs). Second, they spread malware indiscriminately via Japanese P2P (peer-to-peer) file-sharing sites. Third, they specifically target business executives who are traveling overseas and staying at hotels in a number of countries: using a two-step infection process, the attackers first identify their victims and then download further malware to the computers of more significant targets, designed to steal confidential data from the infected computer.2. Our homes and other vulnerabilities
Exploiting unpatched vulnerabilities remains one of the key mechanisms used by cybercriminals to install malicious code on victims' computers. This relies on the existence of vulnerabilities in widely-used software and the failure of individuals or businesses to patch applications.
This year vulnerabilities were discovered in two widely-used open source protocols, known as 'Heartbleed' and 'Shellshock' respectively. Heartbleed, a flaw in the OpenSSL encryption protocol, lets an attacker read the contents of the memory, and intercept personal data, on systems using vulnerable versions of the protocol. OpenSSL is widely-used to secure Internet-based communications, including web, e-mail, instant messaging and Virtual Private Networks (VPN), so the potential impact of this vulnerability was huge. As often happens when there's a risk that personal data might have been exposed, there was a rush to change passwords. Of course, this could only be effective once an online provider had taken steps to patch OpenSSL and thereby secure their systems – otherwise any new password would be just at risk from attackers trying to exploit the vulnerability. We offered some perspectives on the impact of the flaw two months after its disclosure.
In September, the information security world faced a red alert following the discovery of the Shellshock vulnerability (also known as 'Bash'). The flaw allows an attacker to remotely attach a malicious file to a variable that is executed when the Bash command interpreter is invoked (Bash is the default shell on Linux and Mac OS X systems). The high impact of this vulnerability, coupled with the ease with which it could be exploited, caused considerable concern. Many people compared it to Heartbleed. However, unlike Heartbleed, Shellshock provided full system control – not just the ability to steal data from the memory. It didn't take long for attackers to try and take advantage of the vulnerability – we discussed some early examples soon after it was discovered. In most cases attackers remotely attacked web servers hosting CGI (Common Gateway Interface) scripts that have been written in Bash or pass values to shell scripts. However, it remains possible that the vulnerability could have an impact on a Windows-based infrastructure. Unfortunately, the problem wasn't confined only to web servers. Bash is widely used in the firmware of devices that now take for granted in our everyday lives. This includes routers, home appliances and wireless access points. Some of these devices can be difficult or impossible to patch.
The Internet is becoming woven into the fabric of our lives – literally, in some cases, as connectivity is embedded into everyday objects. This trend, known as the 'Internet of Things', has attracted more and more attention. It can seem very futuristic, but the Internet of Things is actually closer than you may think. The modern home today is likely to have a handful of devices connected to the local network that aren't traditional computers – devices such as a smart TV, a printer, a games console, a network storage device or some kind of media player/satellite receiver.
One of our security researchers investigated his own home, to determine whether it was really cyber-secure. He looked at several pieces of household kit, including network-attached storage (NAS) devices, smart TV, router and satellite receiver, to see if they were vulnerable to attack. The results were striking. He found 14 vulnerabilities in the network-attached storage devices, one in the smart TV and several potentially hidden remote control functions in the router. You can read the full details here. It's important that we all understand the potential risks associated with using network devices – this applies to individuals and businesses alike. We also need to understand that our information is not secure just because we use strong passwords or run software to protect against malicious code. There are many things over which we have no control, and to some degree we are in the hands of software and hardware vendors. For example, not all devices include automated update checks – sometimes consumers are required to download and install new firmware. This is not always an easy task. Worse still, it's not always possible to update a device (most devices investigated during this research had been discontinued more than a year before).3. The continuing exponential growth of mobile malware
We have seen dramatic growth in the numbers of mobile malware in recent years. In the period from 2004-13 we analyzed almost 200,000 mobile malware code samples. In 2014 alone we analyzed a further 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 we had 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 (see Mobile Cyber Threats, Kaspersky Lab and INTERPOL Joint Report, October 2014).
53% of all mobile malware detections are now related to malware capable of stealing money. One of the more notable examples is Svpeng, designed to steal money from customers of three of Russia's biggest banks. The Trojan waits until a customer opens an online banking app and replaces it with its own, to try and obtain the customer's login details. It also tries to steal credit card data by displaying its own window over the Google Play app and asking for card details. Another is Waller which, in addition to behaving like a typical SMS Trojan, steals money from QIWI wallets on infected devices.
Cybercriminals have also diversified their efforts to make money from their victims, using methods that have been well-established on desktops and laptops. This includes ransomware Trojans. Fake anti-virus apps are another example of an established approach now being applied to mobile devices. Finally, this year saw the appearance of the first Trojan that is managed through a C2 server hosted in the Tor network. The Torec backdoor is a modification of the commonly-used Tor client, Orbot. The benefit, of course, is that the C2 server can't be shut down.
Until recently, nearly all malware targeting iOS was designed to exploit 'jailbroken' devices.
However, the recent appearance of the 'WireLurker' malware has shown that iOS is not immune from attack.
Mobile devices are now integrated into the fabric of our lives, so it's hardly surprising that the development of mobile malware is underpinned by a cybercrime business that includes malware writers, testers, app designers, web developers and botnet managers.4. Your money or your file(s)
The number of ransomware programs has been growing in recent years. Some simply block access to the victim's computer and demand a ransom payment in order to restore normal access. But many go further than this, encrypting data on the computer. One recent example is 'ZeroLocker'. ZeroLocker encrypts nearly all the files on the victim's computer and adds the extension '.encrypt' to encrypted files (although it doesn't encrypt files located in directories containing the words 'Windows', 'WINDOWS', 'Program Files', 'ZeroLocker' or 'Destroy' and doesn't encrypt files larger than 20MB in size). The Trojan uses a 160-bit AES key to encrypt files. Once the files are encrypted, it runs the 'cipher.exe' utility to remove all unused data from the drive. Both these things make file recovery very difficult. The cybercriminals behind ZeroLocker demand an initial $300 worth of Bitcoins to decrypt the file. If the victim does not pay promptly the fee increases to $500 and $1,000 as time goes on.
Another ransomware program that we analyzed this year is Onion. Not only does this Trojan use the Tor network to hide its C2 servers, but it also supports full interaction with Tor without any input from the victim. Other programs like this communicate with the Tor network by launching (sometimes by injecting code into other processes) the legitimate 'tor.exe' file. By contrast Onion implements this communication as part of the malware code itself. Onion also uses an unorthodox cryptographic algorithm that makes file decryption impossible, even if traffic between the Trojan and the C2 server is intercepted. This Trojan not only uses asymmetric encryption, it also uses a cryptographic protocol known as ECDH (Elliptic Curve Diffie-Hellman). This makes decryption impossible without the master private key – which never leaves the cybercriminals' controlled server.
This year the use of ransomware programs has been extended to devices running Android. The first version of Svpeng, for example, discovered early in 2014, blocks the phone, claiming that the victim was viewing child pornography and demanding a 'fine' of $500 to unlock the phone. A subsequent modification of this malware, discovered in June 2014, completely blocks the device, so that it can only be turned off by pressing down the 'Off' button for a long time – and the Trojan loads again as soon as the device has been switched on again. This version was aimed mainly at victims in the US, but we also saw victims in the UK, Switzerland, Germany, India and Russia. This version demands a payment of $200 to unblock the phone, payment to be made using MoneyPak vouchers. The ransom demand screen displays a photograph of the victim, taken using the frontal camera. Another Trojan, called 'Koler', discovered in May 2014, uses the same approach – blocking access to the device and demanding a ransom payment of between $100 and $300 to unblock the phone. Like Svpeng, this Trojan displays a message claiming to be from the police – it targets victims in more than 30 countries around the world, using local 'police' messages.
Koler's distribution infrastructure
The first Android Trojan to encrypt data, called 'Pletor', appeared in May 2014. This Trojan uses the AES encryption algorithm to encrypt the contents of the phone's memory card and then displays a ransom demand on the screen, payable using the victim's QIWI Visa wallet, MoneXy or standard transfer of money to a telephone number. This Trojan mainly targets victims in Russia and Ukraine (although we have seen victims in other former Soviet republics) and demands the equivalent of around $300 in rubles or hryvnia.
Ransomware operations rely on their victims paying up. Don't do it! Instead, make regular backups of your data. That way, if you ever fall victim to a ransomware program (or a hardware problem that stops you accessing your files) you will not lose any of your data.5. Cha-ching! Using malware to get money from ATMs
Malware for ATMs is not new. The first malware of this kind, called 'Skimer', was found in 2009 – this targeted ATMs in Eastern Europe running a Windows-based operating system. This used undocumented functions to print details of cards inserted in the infected machine and to open cassettes using a master card command. We saw further ATM malware in Brazil, in 2010 ('SPSniffer'): this collected PIN numbers in outdated ATMs using PIN pads that weren't using strong cryptographic protection. Then last year we saw a further family of ATM malware ('Atmer'), designed to steal money from ATMs in Mexico.
This year, at the request of a financial institution, we carried out a forensic investigation into a new attack on ATMs in Asia, Europe and Latin America. The operation was in two stages. The cybercriminals gain physical access to the ATMs and use a bootable CD to install the malware, called 'Tyupkin'; then they reboot the machine to load the malware, putting them in control of the ATM. The malware then runs in an infinite loop, waiting for a command.
To make the scam less obvious, the malware only accepts commands at specific times on Sunday and Monday nights. The attackers can then enter a combination of digits on the ATM keyboard, make a call to the malware operators, enter a further set of numbers and then collect the cash dispensed by the ATM.
Video Footage obtained from security cameras at the infected ATMs showed the methodology used to access cash from the machines. A unique digit combination key based on random numbers is freshly generated for every session: this ensures that no one outside the gang can accidentally profit from the fraud. Then the malicious operator receives instructions by phone from another member of the gang who knows the algorithm and is able to generate a session key based on the number shown: this ensures that the mules collecting the cash do not try to go it alone. When the correct key is entered, the ATM shows how much money is available in each cash cassette, inviting the operator to choose which cassette to rob. Then it dispenses 40 bank notes at a time from the chosen cassette.
The upswing in ATM attacks in recent years is a natural evolution from the more well-established method of using physical skimmers to capture data from cards used in ATMs that have been tampered with. Unfortunately, many ATMs run operating systems with known security weaknesses. This makes physical security even more important; and we would urge all banks to review the physical security of their ATMs.6. Windows XP: forgotten but not gone?
Support for Windows XP ended on 8 April: this means no new security updates, no security hotfixes, free or paid assisted support options or online technical content updates. Sadly, there are still a lot of people running Windows XP – our data suggests that Windows XP accounts for around 18% of infections. This is a lot of people wide open to attack now that security patches have dried up. Effectively, every vulnerability discovered since April is a zero-day vulnerability – that is, one for which there is no chance of a patch. This problem will be compounded as application vendors stop developing updates for Windows XP. Every unpatched application will become yet another potential point of compromise, further increasing the potential attack surface. In fact, this process has already started: the latest version of Java no longer supports Windows XP.
Every Windows XP vulnerability discovered since April is a zero-day vulnerability #KLReportTweet
It might seem that the simple and obvious solution is to upgrade to a newer operating system. But even though Microsoft gave plenty of notice about the end of support, it's not difficult to see why migration to a new operating system might be difficult for some businesses. On top of the cost of switching, it may also mean investing in new hardware and even trying to replace a bespoke application developed specifically for the company – one that will not run on a later operating system. So it's no surprise see some organizations paying for continued XP support.
Of course, an anti-virus product will provide protection. But this only holds good if by 'anti-virus' we mean a comprehensive Internet security product that makes use of proactive technology to defend against new, unknown threats – in particular, functionality to prevent the use of exploits. A basic anti-virus product, based largely on signature-based scanning for known malware, is insufficient. Remember too that, as times goes by, security vendors will implement new protection technologies that may well not be Windows XP-compatible.
Anyone still running Windows XP should see this as a stop-gap, while they finalize a migration strategy. Malware writers will undoubtedly target Windows XP while significant numbers of people continue to run it, since an un-patched operating system will offer them a much bigger window of opportunity. Any Windows XP-based computer on a network offers a weak point that can be exploited in a targeted attack on the company – if compromised this will become a stepping-stone into the wider network.
There's no question that switching to a newer operating system is inconvenient and costly - for individuals and businesses. But the potential risk of using an increasingly insecure operating system is likely to outweigh the inconvenience and cost.7. Beneath the layers of the onion
Tor (short for The Onion Router) is software designed to allow someone to remain anonymous when accessing the Internet. It has been around for some time, but for many years was used mainly by experts and enthusiasts. However, use of the Tor network has spiked this year, in large part because of growing concerns about privacy. Tor has become a helpful solution for those who, for any reason, fear surveillance and the leakage of confidential information. However, our investigations highlighted the fact that Tor is also attractive for cybercriminals, who value the anonymity it offers.
We started seeing cybercriminals actively using Tor to host their malicious infrastructure in 2013. In addition to malware, we found many related resources, including C2 servers, administration panels and more. By hosting their servers in the Tor network, cybercriminals make them harder to identify, blacklist and eliminate. There's also a Tor-based underground marketplace, including the buying and selling of malware and stolen personal data – typically paid for using the crypto-currency Bitcoin, enabling cybercriminals to remain untraceable. Tor allows cybercriminals to conceal the operation of the malware they use, to trade in cybercrime services and launder their illegal profits.
In July we published our analysis of a ransomware Trojan, called 'Onion' that broke new ground in its use of Tor.
Developers of Android-based malware have also started to use Tor. The Torec Trojan, a malware variation of the popular Orbot Tor client, uses a domain in the .onion pseudo zone as a C2 server. Some modifications of the Pletor ransomware Trojan also use the Tor network to communicate with the cybercriminals managing the scam.
Cybercriminals can't always operate with impunity, despite using Tor, as demonstrated by the recent global law enforcement operation against a number of Tor-based cybercrime services ('Operation Onymous').
This begs the question of how the police agencies involved were able to compromise a supposedly 'impenetrable' network – because, in theory at least, there's no way of knowing the physical location of a web server behind a hidden service that someone visits. However, there are ways to compromise a hidden service that don't involve attacking the Tor architecture itself, as we discussed here. A Tor-based service can only remain secure if it's properly configured, if it's free from vulnerabilities or configuration errors and the web application doesn't have any flaws.8. The good, the bad and the ugly
Unfortunately, software isn't neatly divided between good and bad programs. There's always the risk that software developed for legitimate purposes might be misused by cybercriminals. At the Kaspersky Security Analyst Summit 2014 in February we outlined how improper implementation of anti-theft technologies residing in the firmware of commonly used laptops and some desktop computers could become a powerful weapon in the hands of cybercriminals. Our research started when a Kaspersky Lab employee experienced repeated system process crashes on one of his personal laptops, related to instability in modules belonging to the Computrace software developed by Absolute Software. Our colleague hadn't installed the software and didn't even know it was present on the laptop. This caused us concern because, according to an Absolute Software white paper, the installation should be done by the owner of the computer or their IT service. On top of this, while most pre-installed software can be permanently removed or disabled by the owner of the computer, Computrace is designed to survive a professional system cleanup and even a hard disk replacement. Moreover, we couldn't simply dismiss this as a one-off occurrence because we found similar indications of Computrace software running on personal computers belonging to some of our researchers and some enterprise computers. As a result, we decided to carry out an in-depth analysis.
When we first looked at Computrace, we mistakenly thought it was malicious software, because it uses so many tricks that are popular in current malware. Indeed, in the past this software has been detected as malware although at present most anti-malware companies whitelist Computrace executables.
In our view, strong authentication and encryption must be built into a powerful legal surveillance tools #KLReportTweet
We believe that Computrace was designed with good intentions. However, our research shows that vulnerabilities in the software could allow cybercriminals to misuse it. In our view, strong authentication and encryption must be built into such a powerful tool. We found no evidence that Computrace modules had been secretly activated on the computers we analyzed. But it's clear that there are a lot of computers with activated Computrace agents. We believe that it's the responsibility of manufacturers, and Absolute Software, to notify these people and explain how they can deactivate the software if they don't wish to use it. Otherwise, these orphaned agents will continue to run unnoticed and will provide opportunities for remote exploitation.
In June, we published the results of our research into a piece of 'legal' software called Remote Control System (RCS) developed by the Italian company HackingTeam. We discovered a feature that can be used to fingerprint its C2 servers. This allowed us to scan the entire IPv4 space and find all the IP addresses of RCS C2 servers across the globe. We found 326 in total, the greatest number of them located in the US, Kazakhstan and Ecuador. Several IPs were identified as 'government'-related, based on their WHOIS information. Of course, we can't be sure that the servers located in a specific country are being used by law enforcement agencies in that country, but this would make sense: after all, it would avoid cross-border legal problems and avoid the risk of servers being seized by others. We also found a number of mobile malware modules coming from HackingTeam, for Android, iOS, Windows Mobile and BlackBerry. They are all controlled using the same configuration type – a good indication that they are related and belong to the same product family. Unsurprisingly, we were particularly interested in those relating to Android and iOS, because of the popularity of those platforms.
The modules are installed using infectors – special executables for either Windows or Mac OS that run on already-infected computers. The iOS module supports only 'jailbroken' devices. This does limit its ability to spread, but the method of infection used by RCS means that an attacker can run a jailbreaking tool (such as Evasi0n) from the infected computer to which the phone is connected – as long as the device isn't locked. The iOS module allows an attacker to access data on the device (including e-mail, contacts, call history, cached web pages), to secretly activate the microphone and to take regular camera shots. This gives complete control over the whole environment in and around a victim's computer.
We seek to detect and remediate any malware attack, regardless of its origin or purpose #KLReportTweet
The Android module is protected by the DexGuard optimizer/obfuscator, so it was difficult to analyze. But we were able to determine that it matches the functionality of the iOS module, plus offering support for hijacking information from the following applications: 'com.tencent.mm', 'com.google.android.gm', 'android.calendar', 'com.facebook', 'jp.naver.line.android' and 'com.google.android.talk'.
This new data highlighted the sophistication of such surveillance tools. Our policy in relation to such tools is very clear. We seek to detect and remediate any malware attack, regardless of its origin or purpose. For us, there's no such thing as 'right' or 'wrong' malware; and we've issued public warnings about the risks of so-called 'legal' spyware in the past. It's imperative that these surveillance tools don't fall into the wrong hands – that's why the IT security industry can't make exceptions when it comes to detecting malware.9. Privacy and security
The ongoing tension between privacy and security has continued to make headlines.
Among the usual steady stream of security breaches this year, it's not really surprising that the incident that attracted most attention was the theft and subsequent publication of explicit photographs of various Hollywood celebrities. This story highlights the dual responsibility of providers and individuals in securing data stored online. It seems that the theft was made possible by a loophole in iCloud security: the 'Find My iPhone' interface lacked any limitation on the number of password attempts, allowing attackers to brute-force the passwords of the victims. Apple closed up this loophole soon afterwards. However, the attack would not have been possible had the victims not used weak passwords. We increasingly live our lives online. But many of us fail to consider the implications of storing personal data online. The security of a cloud service depends on the provider. The moment we entrust our data to a third-party service, we automatically lose some control over it. It's important to cherry-pick the data we store in the cloud and decide what data is automatically moved from our devices to the cloud.
The issue of passwords is one that keeps surfacing. If we choose a password that is too easy to guess, we leave ourselves wide open to identify theft. The problem is compounded if we recycle the same password across multiple online accounts – if one account is compromised, they're all at risk! This is why many providers, including Apple, Google and Microsoft, now offer two-factor authentication – i.e. requiring customers to enter a code generated by a hardware token, or one sent to a mobile device, in order to access a site, or at least in order to make changes to account settings. Two-factor authentication certainly enhances security – but only if it's required, rather than just being an option.
Two-factor authentication enhances security – but only if it's required, rather than being an option #KLReportTweet
There's always a trade-off between security and ease of use. In an effort to strike this balance, Twitter recently launched its Digits service. Customers no longer need to create a username and password combination in order to sign in to an app. Instead, they simply enter their phone number. They receive a one-time passcode to confirm each transaction – this code is read automatically by the app. Twitter is effectively making itself a go-between, verifying the identity of the customer for the app provider. There are several benefits. Consumers no longer have to worry about creating a login and password combination to set up an account with an app provider; and they don't need to have an e-mail address. App developers don't need to create their own framework for verifying logins; and they won't lose potential customers who don't use e-mail. Twitter gets more visibility into what its customers are interested in. In addition, the fact that no passwords are stored on the app provider's server is also a plus: a breach of an app provider's server will not result in the loss of personal data belonging to customers. However, if someone loses their device, or if it's stolen, the number verification will still work – and anyone with access to the device will be able to access an app in the same way as the legitimate owner. That said, it doesn't represent a step backwards in security compared to the traditional username and password method. Currently, mobile apps don't force a login each time an app is run anyway, so if someone steals a phone, and the owner isn't using a PIN, passcode or fingerprint, the thief has access to everything – e-mail, social networks and apps. In other words, security is dependent on a single-point-of-failure – the PIN, passcode or fingerprint used to access the device itself.
In response to increasing concerns about privacy, the developers of the 'pwnedlist.com' web site created an easy to use interface where people can check to see if their e-mail addresses and passwords have been stolen and published online. This year they have made this a chargeable service.
The response of both Apple and Google to growing fears about loss of privacy was to enable default encryption of data on iOS and Android devices, something that some law enforcement agencies believe plays into the hands of cybercriminals – making it easier for them to evade detection.10. International law enforcement: co-operation brings results
Cybercrime has become an established part of life, on the back of the ever-increasing online activities we engage in. It's tempting to imaging that cybercriminals are able to operate with impunity, but the actions of law enforcement agencies can have a significant impact on their activities. International co-operation is particularly important, given the global nature of cybercrime. This year there have been some notable police successes.
In June 2014 an operation involving law enforcement agencies of several countries, including the UK's NCA (National Crime Agency) and the FBI, was able to take down the global network of computers responsible for managing the 'GameoverZeus' botnet. The police operation ('Operation Tovar') disrupted the communications underlying the botnet, thereby preventing the cybercriminals from controlling it. GameoverZeus was one of the largest operating botnets based on the code of the Zeus banking Trojan. In addition to infecting computers with the Zeus Trojan and stealing login credentials for online e-mail accounts, social networks, online banking and other online financial services, the botnet also distributed the 'Cryptolocker' ransomware program. The police campaign offered victims a breathing-space in which to clean their computers.
Earlier this year Kaspersky Lab contributed to an alliance of law enforcement and industry organizations, co-ordinated by the NCA, to disrupt the infrastructure behind the 'Shylock' Trojan. The Shylock banking Trojan, so-called because its code contains excerpts from Shakespeare's The Merchant of Venice, was first discovered in 2011. Like other well-known banking Trojans Shylock is a man-in-the-browser attack designed to steal banking login credentials from the computers of bank customers. The Trojan uses a pre-configured list of target banks, located in different countries around the world.
In November, Operation Onymous resulted in the take-down of dark markets running within the Tor network.
Recently, an interesting malicious sample was uploaded to a multi-scanner service. This immediately triggered our interest because it appears to represent a previously unknown piece of a larger puzzle. That puzzle is "Turla", one of the most complex APTs in the world.
We have written previously about the Turla APT with posts about their Epic Turla operations and Agent.btz inspiration . So far, every single Turla sample we've encountered was designed for the Microsoft Windows family, 32 and 64 bit operating systems. The newly discovered Turla sample is unusual in the fact that it's the first Turla sample targeting the Linux operating system that we have discovered.
This newly found Turla component supports Linux for broader system support at victim sites. The attack tool takes us further into the set alongside the Snake rootkit and components first associated with this actor a couple years ago. We suspect that this component was running for years at a victim site, but do not have concrete data to support that statement just yet.
The Linux Turla module is a C/C++ executable statically linked against multiple libraries, greatly increasing its file size. It was stripped of symbol information, more likely intended to increase analysis effort than to decrease file size. Its functionality includes hidden network communications, arbitrary remote command execution, and remote management. Much of its code is based on public sources.Md5 Size Verdict Name 0994d9deb50352e76b0322f48ee576c6 627.2 kb N/A (broken file) 14ecd5e6fc8e501037b54ca263896a11 637.6 kb HEUR:Backdoor.Linux.Turla.gen
General executable characteristics:
ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), statically linked, for GNU/Linux 2.2.5, stripped
Statically linked libraries:
- glibc2.3.2 - the GNU C library
- openssl v0.9.6 - an older OpenSSL library
- libpcap - tcpdump's network capture library
Hardcoded C&C, known Turla activity: news-bbc.podzone[.]org
The domain has the following pDNS IP: 220.127.116.11
Note: the C&C domain is currently sinkholed by Kaspersky Lab.
The sample is a stealth backdoor based on the cd00r sources.
This Turla cd00r-based malware maintains stealth without requiring elevated privileges while running arbitrary remote commands. It can't be discovered via netstat, a commonly used administrative tool. It uses techniques that don't require root access, which allows it to be more freely run on more victim hosts. Even if a regular user with limited privileges launches it, it can continue to intercept incoming packets and run incoming commands on the system.Startup and Execution
To start execution, the process requires two parameters: ID (a numeric value used as a part of the "magic packet for authentication") and an existing network interface name. The parameters can be inputted two different ways: from STDIN, or from dropper a launching the sample. This is NOT a command-line parameter, it's a real prompt asking the attacker user to provide the input parameters. After the ID and interface name are entered and the process launched, the backdoor's process PID is returned. Here is a screenshot of this simple interface:
While there is no initial network callback, a section of code maintains a hardcoded c2 string "news-bbc.podzone[.]org". This fully qualified domain name was first set up in 2010, suggesting that this binary is fairly recent in the string of Turla campaigns. Also, while we haven't seen additional file download activity from this server by this tool, it likely participated as a file server of sorts.Magic Packets for Remote Command Execution
The module statically links PCAP libraries, and uses this code to get a raw socket, applies a filter on it, and captures packets, checking for a specific condition (the *original cd00r first used this method, based on ports and SYN-packets). This condition is expressed here (it is based on the ID value input at startup by the attacker):ID = 123 Filter = (tcp[8:4] & 0xe007ffff = 0xe003bebe) or (udp[12:4] & 0xe007ffff = 0xe003bebe) ID = 321 Filter = (tcp[8:4] & 0xe007ffff = 0x1bebe) or (udp[12:4] & 0xe007ffff = 0x1bebe)
In simple terms, it checks for an ACK number in the TCP header, or the second byte from the UDP packet body.
If such a packet is received and the condition check is successful, execution jumps to the packet payload contents, and it creates a regular socket. The backdoor handles this socket as a file with read/write operations. It's not the typical recv/send used in this code. It uses this new socket to connect to the source address of the "magic packets". Then it reports its own PID and IP to the remote address, and starts an endless loop for receiving remote commands. When a command arrives, it is executed with a "/bin/sh -c " script.Unused Code
The sample contains much unused code (at least, it isn't called directly from within the backdoor's code). Also, this sample has duplicate code, for instance - getting IP interface. This unfinished programming again tells us the sample is similar to a debug or beta version.
Much of the unused code is related to file I/O:
- parsing('/root/.tmpware') – parsing, syntax check (begin, end, table) – reads (fscanf) file by format "%*s %o %s", where %o %s is mode and pathname. Then these parameters are passed to chmod(pathname, mode). E.g. sample changes format of files, specified in some other structured file.
chmod(fd, 140h) – only owner
The descriptor then is used everywhere in the code for reading and writing various files. The sample writes to available file paths:PATH=/bin:/usr/bin:/usr/local/bin:/usr/openwin/bin:/usr/ucb/bin:/ usr/ccs/bin
Although Linux variants from the Turla framework were known to exist, we haven't seen any in the wild yet.
This specific module appears to have been put together from public sources with some added functionality from the attackers. Some of the malicious code appears to be inactive, perhaps leftovers from older versions of the implant. Perhaps the most interesting part here is the unusual command and control mechanism based on TCP/UDP packets, as well as the C&C hostname which fits previously known Turla activity.
The discovery of this Turla module rises one big question: how many other unknown Turla variants exist?
Update: Since the publishing of this blogpost, we have discovered another Linux Turla module, which apparently represents a different malware generation than the previously known samples:
The new sample was heuristically detected by our product due to similarities with the previously discovered samples.Md5 Size Verdict Name 19fbd8cbfb12482e8020a887d6427315 801,561 bytes HEUR:Backdoor.Linux.Turla.gen Related research:
- BAE Systems - The Snake Campaign
- Kaspersky Lab - The Epic Turla Operation
- "TR-25 Analysis - Turla / Pfinet / Snake/ Uroburos" by CIRCL.LU
- "Uroburos: the snake rootkit", technical analysis by deresz and tecamac
- Agent.BTZ - A Source of Inspiration?