Threat Intelligence

Hunting Malware in the Deep Web

Pierluigi Paganini
February 10, 2015 by
Pierluigi Paganini


Botnets are still considered one of the most dangerous cyber threats. These malicious networks of compromised machines are used by cyber criminals and state-sponsored hackers for numerous activities, including DDoS attacks, spam campaigns, and financial scams.

The principal problem for a botmaster is to make a botnet resilient against operations run by law enforcement. For operators it is essential to hide Command and Control servers and network traffic to avoid takeover of the malicious infrastructure. The Tor network offers a privileged environment for botmasters that could exploit the popular anonymizing network to hide the C&C servers.

Hands-on threat intel training

Hands-on threat intel training

Learn how to collect, analyze and act on cyber threat intelligence with expert instruction and hands-on exercises in Infosec Skills.

Tor botnets

During the Defcon Conference in 2010, security engineer Dennis Brown discussed Tor-based botnets, highlighting pro and cons of the choice to hide C&C servers in the Tor network. The principal advantages of Tor-based botnets are:

  • Availability of Authenticated Hidden Services
  • Availability of Private Tor Networks
  • Possibility of Exit Node Flooding

Security researchers use traffic analysis to detect botnet activities and to localize the C&C servers. Typically they do this by using Intrusion Detection Systems and network analyzers. Once they've detected a botnet, the researchers and law enforcement have different options to eradicate it:

  • Obscuration of the IP addresses assigned to the C&C server
  • Cleaning of server hosting botnet and of the compromised hosts
  • Domain name revoke
  • Hosting provider de-peered

The botnet traffic is routed to the C&C server through the Tor network that encrypts it, making its analysis more difficult.

Brown proposed the following two botnet models that exploit the Tor network:

  • "Tor2Web proxy based model"
  • "Proxy-aware malware over Tor network"

"Tor2Web proxy based model"

The routing mechanism relies on the Tor2Web proxy to redirect .onion web traffic. The bot has to connect to the hidden service passing through the Tor2Web proxy pointing to an onion address that identifies the C&C server that remains hidden. The principal problem related to this approach is that it is easy to filter Tor2Web traffic, and a similar configuration could suffer from considerable latencies due to the Tor network that could make a botnet built with this approach unresponsive.

"Proxy-aware Malware over Tor network"

This approach is based on making use of proxy-aware malware. Due to the absence of the Tor2Web service, the bot agents have to run Tor clients on the infected hosts. The main difference with respect to the first solution is in the requirements for the bot agents and their configuration. Bots need to have SOCKS5 support to reach .onion addresses through the Tor network by loading Tor on the victims' systems.

This second approach is more secure because traffic isn't routed through a proxy and it is entirely within the Tor network due the direct connection between Bots and C&C servers. This configuration avoids traffic interception from exit nodes that are not involved in the architecture.

This approach is more complex from a Bot perspective due to the complexity in managing the SOCKS5 interface and in botnet synchronization. This kind of botnet could be easily detected by the presence of Tor traffic on a network.

Strengths and weaknesses of Tor botnets

Among the strengths:

  • Botnet traffic masquerades as legitimate Tor traffic
  • Encryption prevents most Intrusion Detection Systems from finding botnet traffic
  • P2P architecture makes botnets more resilient to take down
  • Difficulty for the localization of the command and control servers (C&C)
  • Hidden Services provide a Tor-specific .onion pseudo top-level domain, which is not exposed to possible sinkholing.
  • The operator can easily move around the C&C servers just by re-using the generated private key for the Hidden Service.

Among the weaknesses:

  • Complexity of botnet management
  • Risk of botnet fragmentation
  • Latency in the communication

Tor botnets: real cases

The Skynet botnet

One of the first examples of a Tor based botnet is the Skynet botnet that was discovered in December 2012 by experts at G-Data and Rapid7. The bot was a strain of the popular Zeus trojan, which included a Tor client for Windows and a bitcoin mining tool. The researchers at G-Data also reported that Skynet used hidden IRC services with Tor to control the malicious architecture.

The Skynet botnet can fulfill different tasks such as mining bitcoin or providing bot agents to involve in illegal activities such as DDoS attacks or spam campaigns.

Figure 1 - Tor botnet

Mevade botnet

Going forward in time, we find the Mevade botnet (a.k.a Sefnit, LazyAlienBiker). In September 2013 it caused a spike in the number of Tor users, which reached 5 million active users.

Figure 2 - Tor metrics: Mevade spikes Tor users

Authors of Mevade's Tor variant appear to use the Russian language. The purpose of the botnet was the installation of adware and toolbars onto the victim's systems, mine Bitcoin and steal sensitive information from the infected PC. Experts at TrendMicro revealed that the Mavade malware had also a "backdoor component and communicates over SSH to remote hosts" that made the agent ideal for data theft.

The Atrax crimekit

In November 2013, researchers from Danish security firm CSIS discovered a new crimekit, dubbed Atrax, which was sold in the underground market. One of the main features implemented by its authors is the ability to exploit Tor networks to communicate with Command & Control servers. The Atrax crimekit was cheap – it was offered for $250, and among the other features implemented by its authors, there were:

  • Virtual currency mining (Bitcoin mining and Litecoin mining)
  • Browser data extraction
  • Availability of a module to run DDoS attacks that offers complete support for both Full IPv6 and IPv4 and implements principal attack techniques including UDP Flood, TCP Flood, TCP Connect Flood, HTTP Slowloris, and many other methods.
  • Data stealing, including Bitcoin wallets (such as Armory, Bitcoin-Qt, Electrum and Multibit).

Figure 3 - Atrax crimekit

The Atrax crimekit has a modular structure. The malware includes a series of add-ons that implement the functionalities described. A plugin which implements a data stealer was sold for $110, the form grabber runs for $300, and an experimental add-on for coin mining was sold for $140.

It's interesting to note that the Atrax crimekit was sold with free updates, bug fixes and support. Below a list of standard features present in the Atrax crimekit:

  • Kill
  • Update
  • Download (over Tor), Execute (Commandline-Parameter allowed)
  • Download (over Tor), Execute (Commandline-Parameter allowed) in memory
  • Install Plugin
  • Installation List (A list with all installed applications)

64-bit ZeuS banking trojan using Tor network

In December 2013, security researchers at Kaspersky Lab detected a new strain of the popular Zeus trojan. The new variant was designed to operate on 64-bit, and authors enhanced the malicious code with the support of communication through the Tor network.

This version of the popular banking trojan also used a web injection mechanism to steal banking credentials from the victim's browser. It was also able to steal digital certificates and implement a keystrokes feature.

The authors implemented a communication mechanism with the C&C server over the Tor network, a feature that makes it more difficult for law enforcement and security firms to track botnets.

The 64-bit version of the Zeus banking trojan executes a Tor component, starting the svchost application in suspended mode and then injecting the Tor code into that process, running it in a stealth mode. The malicious traffic was routed through TCP port 9050 and the stolen data were sent to the onion domain with address egzh3ktnywjwabxb [.] onion.

"Tor.exe is launched indirectly -- ZeuS starts the system svchost.exe application in suspended mode, then injects the tor.exe code into this suspended svchost.exe process, tunes the code to run properly and resumes execution of the suspended svchost," Tarakanov explains. "As a result, instead of the system svchost.exe, the process actually starts executing tor.exe." states the blog post published on SecureList.

Figure 4 -The Tor utility under the cover of the svchost.exe process creates an HTTP proxy server

Another peculiarity of the malware is that it instantiates a hidden service that creates a configuration file for any victims, which includes a unique private key for the service and an exclusive domain. The feature allows the botmaster to control the architecture via Tor.

"The botnet operator will be aware of the generated onion domain related to every infected machine as the malware informs the CnC about its tor domain name. So, when an infected machine is online the botnet operator can reach it connecting to its unique onion domain via the Tor network. One purpose of this approach is the remote control of the infected host. For example, one of these ports specifically listens to in the VNC function of ZeuS, obviously meaning that ZeuS provides remote desktop control to the operator via this port," continues the post.

This version of the Zeus trojan was able to trigger its execution after one program within a list of 100 predefined applications is started.

ChewBacca financial malware

In early 2014 the researchers at RSA discovered a variant of the banking Trojan ChewBacca that was used to steal credit card data from infected POS systems. Also in this case, the botnet was controlled by servers hidden in the Tor network.

According to the experts at RSA, the botnet based on the ChewBacca POS variant was used against customers in at least 11 countries (including US, Russia, Canada and Australia) since October 25, 2013. The malware was able to steal credit card data with "keylogger" capabilities or dumping the memory content of POS systems in search for credit card details. The bot is able to collect track 1 and track 2 data of payment card during purchases.

"Chewbacca code was compiled with Free Pascal 2.7.1., once executed windows based system, it drops as spoolsv.exe in the startup folder and also drops a copy of Tor"

"After execution, the function "P$CHEWBACCA$_$TMYAPPLICATION_$__$$_INSTALL" is called, which drops itself as "spoolsv.exe" into the "Startup folder" (e.g. C:Documents and SettingsAll UsersStart MenuProgramsStartup) and requests the public IP of the victim via a publicly accessible service at (which is not related to the malware). Tor is dropped as "tor.exe" to the user-s Temp and runs with a default listing on "localhost:9050″."

Figure 5 - ChewBacca console

The Bifrose malware

In August 2014, researchers from TrendMicro detected a new variant of the Bifrose malware leveraging on the Tor network. The new variant of the Bifrose backdoor was used in a targeted attack against a device manufacturer. Bifrose has been around for many years, and it is quite easy to acquire in the underground. The malware has a data stealing ability, but it is mostly popular for its keylogging routines. The variant detected by the malware experts at TrendMicro (detected as BKDR_BIFROSE.ZTBG-A - hash 5e2844b20715d0806bfa28bd0ebcba6cbb637ea1) leverages the Tor network to hide communications between the infected machines and the C&C server.

"What makes this variant more elusive is its ability of Tor to communicate with its command-and-control [C&C] server," reports a blog post published by TrendMicro.

The Bifrose malware was widely used by cyber criminals. In 2010 a threat actor targeted human resource (HR) personnel of different government offices, including the African Union and the NATO. The Bifrose variant used in the targeted attack on the device manufacturer was able to perform the following operations, as explained in the blog post:

  • Download a file
  • Upload a file
  • Get file details (file size, last modified time)
  • Create a folder
  • Delete a folder
  • Open a file using ShellExecute
  • Execute a command line
  • Rename a file
  • Enumerate all windows and their process IDs
  • Close a window
  • Move a window to the foreground

OnionDuke: APT Attacks exploited the Tor Network

In November 2014, the experts from F-Secure discovered a link between the crew operating a rogue Tor node used to spread OnionDuke malware and MiniDuke APT.

Just a month before, the security researcher Josh Pitts of Leviathan Security Group identified a Russian Tor exit node that was patching the binaries downloaded by the users with malware. The expert reported it to officials of the Tor Project, who flagged the Tor exit node as bad and shut down it. Further investigations on the case revealed that the threat actors that managed the node were serving malware through the explained scheme for more than a year.

Figure 7 - OnionDuke infection

The bad actors used the Tor exit node to serve a backdoor, dubbed OnionDuke, to the victim's machine with a man-in-the middle attack in the downloading phase.

Security experts at F-Secure discovered that the rogue exit node was tied to the MiniDuke criminal crew. MiniDuke is the name of a sophisticated cyber espionage campaign discovered in 2013 by experts at Kaspersky Lab and Hungary's Laboratory of Cryptography and System Security (CrySyS). The MiniDuke APT infected dozens of machines at government agencies across Europe. Exploiting a security flaw in Adobe software, the malicious payload is dropped once the victim opens the malicious PDF file.

The malware was used by attackers to steal sensitive data from government and high profile entities. The researchers speculated that the level of sophistication and the nature of the chosen targets suggest that the attacks are part of a state-sponsored espionage campaign.

According to the experts, "OnionDuke," the malware spread through the bogus exit node, is a malware different from the ones used in the past by the threat actors behind the MiniDuke crew.

It must be noted that all five domains contacted by OnionDuke aren't dedicated malicious servers. Instead, they are legitimate websites compromised by threat actors.

The experts identified different samples of the malware and multiple other components of the OnionDuke malware family, which were designed to execute specific tasks like data stealing.

The analysis of the various samples allowed the researchers at F-Secure to discover the link with the MiniDuke gang. The owner of the Command & Control (C&C) server used to control a sample of the OnionDuke backdoor (W32/OnionDuke.A) is the same that was involved in the MiniDuke agent.

This circumstance suggests that although OnionDuke and MiniDuke are two separate strains of malware, the threat actors behind them shared the control infrastructure.

"One component, however, is an interesting exception. This DLL file (SHA1 d433f281cf56015941a1c2cb87066ca62ea1db37, detected asBackdoor:W32/OnionDuke.A) contains among its configuration data a different hardcoded C&C domain, and also evidence suggesting that this component may abuse Twitter as an additional C&C channel. What makes the domain interesting, is it was originally registered in 2011 with the alias of 'John Kasai'. Within a two-week window, 'John Kasai' also registered the following domains:,,,,,,,,,, and This is significant because the domains and have previously been identified as C&C domains used by MiniDuke," reports F-Secure in the blog post.

CryptoWall Ransomware is resurrected with new features

In early 2015, the researchers at Cisco's Talos group published an analysis of a new variant of Cryptowall ransomware that implements a series of new features, including the exploitation of the Tor anonymity network to hide its command-and-control infrastructure.

The new variant of CryptoWall was improved by cyber criminals that applied the necessary modifications to its code to make it resilient to the operation of law enforcement.

Cisco's Talos Security Intelligence and Research Group reported that the new strain of the CryptoWall ramsonware is able to distinguish between 32- and 64-bit architectures and to execute different versions for each and OS, including the newest versions of Mac OS X.

"The latest Cryptowall 2.0, utilizes TOR to obfuscate the command and control channel. The dropper utilizes multiple exploits to gain initial access and incorporates anti-vm and anti-emulation checks to hamper identification via sandboxes. The dropper and downloaded Cryptowall binary actually incorporate multiple levels of encryption. One of the most interesting aspects of this malware sample, however, is its capability to run 64 bit code directly from its 32 bit dropper," states the report.

The attack chain starts with a phishing mail that includes the CryptoWall variant in a ".zip" attachment. The compressed archive included an exploit that relies a Microsoft privilege escalation vulnerability (CVE-2013-3660) to compromise the target machine.

"CryptoWall 2.0 can be delivered through multiple attack vectors, including email attachments, malicious pdf files and even various exploit kits. In the sample that we analyzed, the dropper utilized CVE-2013-3660, 'Win32k.sysElevation of Privilege Vulnerability' to achieve the initial privilege escalation on X86 based machines. This exploit works on 32 bit OSs starting beginning with Vista. The dropper even includes a 64-bit DLL that is able to trigger the exploit in all the vulnerable AMD64 Windows Systems."

This new variant of CryptoWall also implements an anti-VM and anti-emulation check pass that prevents the execution in a virtualized environment for malware analysis.

CryptoWall implements a multistep decryption. In the first phase, it decrypts just a first portion of code to check if it is running in a virtualized environment. If it passes the check, it then continues to decrypt. According to the Cisco researchers, the feature could be exploited to prevent the execution of the malware by adding fake entries in the file system that indicate a virtual machine is running.

Once it has infected the machine, the sample connects to the Tor Servers with an encrypted SSL connection on port 443 or 9090. The C&C servers discovered by the researchers were using the following Tor URLs:

  • crptarv4hcu24ijv.onion
  • crptbfoi5i54ubez.onion
  • crptcj7wd4oaafdl.onion

"Using hardcoded IP address in the PE, the malware connects to the TOR Server with an encrypted SSL connection on port 443 or 9090. After successfully connecting, it starts to generate the Cryptowall domain names using a customized Domain Generation Algorithm (DGA). The algorithm is located at offset + 0x2E9FC."

Citroni ransomware

Recently a security researcher analyzed a new ransomware dubbed Critroni, which is being sold in different underground forums. Critroni (aka CTB-Locker) is the name of a new ransomware that has been recently included in the Angler exploit kit. A detailed analysis of the ransomware was posted on "" by the French security researcher Kafeine.

Critroni implements many functionalities, including the ability to exploit the Tor network to host its command and control.

"Placing a server in onion-domain (TOR), close to domain abuse can not be practically impossible to trace the owner and shut down the server. Connection to the server only after encryption of all files. Early Detection is not possible on the traffic, it is impossible to block the work of the locker. Blocking TOR prevents only payment the user, not the program. Analogs are connected to the server until the crypt and can block," states the ad for the malware.

The experts explained that the success of the Critroni ransomware was advantaged by the takedown of the GameOver Zeus managed by law enforcement last year. The botnet in fact was used by cyber criminals to serve CryptoLocker ransomware.

Around the same time in mid-June, security researchers began seeing advertisements for the Critroni ransomware on underground forums. The malware was sold for around $3,000. The Critroni agent was initially spread exclusively in Russia; later its presence was detected in many other countries worldwide. Many criminal groups are using Citroni for their extortion activities. They used to serve the ransomware as part of the Angler exploit kit, which serves a spambot on victims' machines. The spambot module is used by malware authors to drop a couple of other payloads. One of them is Citroni.

Critroni encrypts a variety of files on the targeted machine and then displays a dialogue box that demands a payment in Bitcoins in order to decrypt the files.

Figure 8 - Citroni ransomware

Victims have to pay the ransom within 72 hours. If they haven't any Bitcoins, the ransomware provides detailed instructions on how to acquire them.

I2P botnet: real cases

Not only Tor network - CryptoWall 3.0 uses I2P network

The Tor network isn't the only anonymizing network exploited by malware authors to hide their malicious infrastructure. In early 2015 a new version of the infamous CryptoWall ransomware was spotted by Microsoft, just a week after the Cisco's Talos Security Intelligence and Research Group announced the discovery of a new strain of the same malware that exploits the Tor network. The new variant of CryptoWall ransomware, like others, is distributed via malicious email and through malvertising campaigns.

This variant was dubbed by the researchers CryptoWall 3.0 or Win32/Crowti, and it isn't so different from previous instances. However, the experts noted that the names of the files containing the ransom demand have been changed to "HELP_DECRYPT."

This variant customizes files for each infected machine and provides victims a personalized link to a page that contains includes instructions. The instruction page is still reached through the Tor network. The victims of the CryptoWall 3.0 are given 7 days to pay $500 in Bitcoins if they want to decrypt their documents, but if they don't pay in 7 days, the ransom increases to $1,000.

On January 12, Microsoft identified 288 unique CryptoWall ver. 3.0 infections.

"The graph below shows the spike after two days of no activity from 288 unique machines affected by this malware," reads the post published Microsoft.

Figure 9 - Cryptowall ver. 3.0 infections

The French researcher Kafeine who analyzed CryptoWall 3.0 reported that the communications to C&C served are encoded with the RC4 cipher. Another feature implemented in the latest variant of the malware is the support of I2P (Invisible Internet Project) for C&C communications.

"It seems communication with the C&C are Rc4 encoded (key seems to bealphanum sorted path of the POST ) and using i2p protocol," said Kafeine.

I2P is another anonymizing network used to hide the location of the control servers and make the botnet resilient the C&C to the law enforcement.

Also recently, a new version of the popular black market Silk Road, Silk Road Reloaded, migrated on I2P, probably because at this moment there is the conviction that it is more secure than Tor.

It happens now … new Dyre banking trojan variant

A few days ago, the experts at TrendMicro spotted a new variant of the DYRE /Dyreza banking malware with new propagation and evasion techniques. The malware is spread through malicious emails containing the Upatre downloader disguised as a fax or the details of a package delivery, but once it is executed, the download drops the new Dyre variant, which in turn downloads the WORM_MAILSPAM.XDP worm.

The propagation technique implemented by the cyber criminals is very effective. The worm exploits the Microsoft Outlook email client present on the victim's machine to spread spam emails with the Upatre downloader attached to them. The emails aren't sent to the victim's contacts, instead they are sent to email addresses passed by the C&C server. Once the emails are sent by the worm, it deletes itself. This variant of Dyre uses hard-coded addresses for its IP addresses. The malware authors also implemented backup mechanisms for command and control infrastructure that rely on a URL provided by the malware's domain generation algorithm (DGA) or a hard-coded address of a C&C server hidden on the Invisible Internet Project (I2P) network.

Figure 10 - Dyre I2P

In this case, the I2P network is used as a supplementary way to control the botnet, a choice to make it more resilient to attacks.


Security experts believe that malware authors will continue to exploit anonymizing networks like Tor and I2P. Analyzing the timeline of malware detections made by principal security firms, cyber criminals have been increasing the adoption of such networks since 2012.

Figure 11 - Malware in the Deep Web (Security Affairs)

Malware authors will exploit the Deep Web basically as a backup mechanism for their botnet and to make them more resistant to various kinds of attacks operated by law enforcement.


Hands-on threat intel training

Hands-on threat intel training

Learn how to collect, analyze and act on cyber threat intelligence with expert instruction and hands-on exercises in Infosec Skills.

Pierluigi Paganini
Pierluigi Paganini

Pierluigi is member of the ENISA (European Union Agency for Network and Information Security) Threat Landscape Stakeholder Group, member of Cyber G7 Workgroup of the Italian Ministry of Foreign Affairs and International Cooperation, Professor and Director of the Master in Cyber Security at the Link Campus University. He is also a Security Evangelist, Security Analyst and Freelance Writer.

Editor-in-Chief at "Cyber Defense Magazine", Pierluigi is a cyber security expert with over 20 years experience in the field, he is Certified Ethical Hacker at EC Council in London. The passion for writing and a strong belief that security is founded on sharing and awareness led Pierluigi to find the security blog "Security Affairs" recently named a Top National Security Resource for US.

Pierluigi is a member of the "The Hacker News" team and he is a writer for some major publications in the field such as Cyber War Zone, ICTTF, Infosec Island, Infosec Institute, The Hacker News Magazine and for many other Security magazines.