- Author: Dell SecureWorks Counter Threat Unit™ Threat Intelligence
- Date: 15 June 2015
Malware authors are evolving their techniques to evade network and host-based detection mechanisms. Stegoloader could represent an emerging trend in malware: the use of digital steganography to hide malicious code. The Stegoloader malware family (also known as Win32/Gatak.DR and TSPY_GATAK.GTK despite not sharing any similarities with the Gataka banking trojan) was first identified at the end of 2013 and has attracted little public attention. Dell SecureWorks Counter Threat Unit(TM) (CTU) researchers have analyzed multiple variants of this malware, which stealthily steals information from compromised systems. Stegoloader's modular design allows its operator to deploy modules as necessary, limiting the exposure of the malware capabilities during investigations and reverse engineering analysis. This limited exposure makes it difficult to fully assess the threat actors' intent. The modules analyzed by CTU researchers list recently accessed documents, enumerate installed programs, list recently visited websites, steal passwords, and steal installation files for the IDA tool.
Stegoloader has a modular design and uses digital steganography to hide its main module's code inside a Portable Network Graphics (PNG) image downloaded from a legitimate website. Other malware families have used this technique, including the Lurk downloader, which CTU researchers analyzed in April 2014. At the end of 2014, CTU researchers also observed the Neverquest version of the Gozi trojan using this technology to hide information on its backup command and control (C2) server.
Stegoloader’s deployment module downloads and launches the main module; it does not have persistence. Before deploying other modules, the malware checks that it is not running in an analysis environment. For example, the deployment module monitors mouse cursor movements by making multiple calls to the GetCursorPos function. If the mouse always changes position, or if it does not change position, the malware terminates without exhibiting any malicious activity.
In another effort to slow down static analysis, most of the strings found in the binary are constructed on the program stack before being used. This standard malware technique ensures that strings are not stored in clear text inside the malware body but rather are constructed dynamically, complicating detection and analysis.
Before executing its main function, Stegoloader lists the running processes on the system and terminates if a process name contains one of the strings in Table 1. Most of the strings represent security products or tools used for reverse engineering. Stegoloader does not execute its main program code if it detects analysis or security tools on the system.
Table 1. Strings causing Stegoloader to terminate.
At every stage of its execution, the deployment module reports its status to a C2 server using HTTP GET requests. Figure 1 shows a trace of reports sent from a compromised system to its C2 server. The GET requests are constructed from a list of preconfigured URLs. In the example shown in Figure 1, the first string after the "report_" substring is the hex-encoded name of the computer where the malware is running. The second substring is a hex-encoded pointer used to list files in the victim's home directory (returned by the FindFirstFileA() function). Appendix A lists the status messages that can be sent by the Stegoloader deployment module.
The deployment module fetches a PNG image from a legitimate hosting website (see Figure 2). The image's URL is hard-coded in the binary. After downloading the image, Stegoloader uses the gdiplus library to decompress the image, access each pixel, and extract the least significant bit from the color of each pixel. The extracted data stream is decrypted using the RC4 algorithm and a hard-coded key. Neither the PNG image nor the decrypted code is saved to disk, making the malware difficult to find via traditional disk-based signature analysis. The image's URL and the RC4 key vary in the samples analyzed by CTU researchers.
After the main Stegoloader module is downloaded and decrypted, the deployment module transfers execution to the main module, which resides in a memory area that has been allocated for this purpose. The deployment module is dormant until the main module finishes executing. When the main module terminates, the deployment module sends a last report to its C2 server indicating the main module has finished, and then it also terminates.
The main Stegoloader module communicates with its C2 server via HTTP POST requests (see Figure 3) and executes commands sent by the malware operator. Communications are encrypted using the RC4 algorithm and a hard-coded 16-byte key. The POST URL is hard-coded in the body of the malware.
Figure 4 shows the decrypted header of an HTTP POST request performed by a compromised system.
- The first 16 bytes (in red) are randomly generated and change with each request.
- The next 16 bytes (in blue) are also randomly generated but are used as a session identifier. They are constant across all messages sent from a compromised system during the same execution of the malware.
- The following byte is always '01'.
- The eight bytes (in green) are copied from the previous response received from the C2 server. They are used as a sequence number but are not incremental. If the compromised system is sending its first message, the bytes are initialized to '0'.
- The byte at offset 41 (in orange) is the command that was previously sent by the C2 server and is the command that is currently being answered. This value is set to zero if the compromised system is contacting its C2 server for the first time.
- The next two bytes are flags. The first byte indicates that there was no error executing the previous operation, and the '01' flag indicates further data will follow the header.
- The last four bytes (in purple) indicate the length for the rest of the message. Messages longer than 400 bytes are compressed using the lzma compression algorithm.
Before sending this message to its C2 server, Stegoloader prepends a CRC32 (cyclic redundancy check) of the message and encrypts the message with its checksum using RC4. The message is then prepended with 16 hard-coded bytes, which are likely associated with the RC4 key used to decrypt the message on the server side.
The C2 server's response is also RC4-encrypted. Figure 5 shows a decrypted message from the C2 server. The first field of the response (in red) is a CRC32 for the rest of the message. The command code (in blue) is located at offset 29 (0x1d). The bytes between offsets '21' and '28' (in green) are used as a session identifier and are returned to the server in the next C2 request. Additional code or data sent by the C2 server is appended after these first 32 bytes.
Table 2 lists commands that can be executed by Stegoloader.
|Command code||Command description|
|0x01||Kill command, stops execution|
|0x02||Save content from HTTP response to temporary location and execute it by calling the CreateProcessA function from kernel32.dll|
|0x03||Send system information (see Appendix B)|
|0x04||Send list of software installed on the compromised system (determined by enumerating registry keys used to uninstall software)|
|0x05, 0x06, 0x07||Send Firefox, Chrome, and Internet Explorer browser history to the C2 server|
|0xDC, 0xDD, 0xDE, 0xDF||Sleep command, no operation is performed, ping C2 server again after 3 minutes|
|0x64||Execute shellcode (shellcode is passed directly after the message header)|
Table 2. Stegoloader command codes and descriptions.
The main Stegoloader module gathers information about compromised systems. If the information matches specific criteria, the malware operator can deploy additional modules. Table 3 lists hashes of additional modules discovered by CTU researchers. These modules are directly executed in memory and are never saved to disk.
|SHA1 hash||Size (in bytes)||Module name|
|4dedc828d835ae6efa5740fcb640bf010303d02d||7312||List recently opened documents|
|ce354abcaa7143ea4de30d69da2edc9d359f8f2c||38407||Pony password stealer|
Table 3. Additional Stegoloader modules in samples analyzed by CTU researchers.
The IDA interactive disassembler is frequently used by reverse engineers and malware analysts to analyze malicious software. Stegoloader has a module that steals installed instances of the IDA software. If IDA is detected on the compromised system, the C2 server sends and executes this module. This module uses a different C2 server than the main Stegoloader module. Its reporting pattern is very similar to the deployment module.
The IDA-stealing module searches for IDA-related entries in the registry and sends discovered files to a legitimate online file-hosting website. Table 4 lists files that may be exfiltrated. When a file is uploaded, the file-hosting website returns a link that can be shared with users who want to download the file. The malware parses the response and sends the download links for the exfiltrated files to its C2 server.
Table 4. Files uploaded to the file-hosting website if discovered on a compromised system.
List recently opened documents
The module to list recently opened documents uses the SHGetFolderPathA function from shell32.dll with the CSIDL_RECENT parameter to find the system folder used to store links for the victim's most recently used documents. The module parses the links, resolves their location on the local hard drive, and sends the list of recently used links and files to the C2 server using the same server and protocol as the main module.
Host geographic localization
The host geographic localization module starts an Internet Explorer instance and visits two web pages, ip2location.com and whoer.net, which return information about the visitor's public-facing IP address. The websites also include geolocation information for the visiting IP address. The module then compresses and returns the HTML content to its C2 server using the same server and protocol as the main module.
Pony password stealer
Stegoloader's Pony password stealer module is a copy of the Pony Loader information stealing malware. Since the leak of Pony Loader's source code on underground forums at the end of 2013, it has been used in various operations. This module can steal passwords for most popular applications used for protocols such as POP, IMAP, FTP, and SSH. The information stolen by the Pony password stealer module is packaged and sent to the main module's C2 server using the same protocol as the main module.
Additional tactics, techniques, and procedures
The oldest Stegoloader samples located by CTU researchers were submitted to VirusTotal at the end of 2013. Variants have used filenames related to software piracy. One sample, which used Avanquest_PowerDesk_9_0_1_10_keygen.exe, was bundled with software piracy software that was executed at the same time as the Stegoloader deployment module (see Figure 6).
Dell SecureWorks data indicates that this malware family has affected multiple verticals, including healthcare, education, and manufacturing. The malware has the characteristics of a stealthy and opportunistic information stealer. It has not been observed being used with exploits or spearphishing, making it more similar to "mass market" commodity malware than to a tool used in targeted attacks.
Some Stegoloader variants have been observed downloading and installing the Vundo (also known as Ponmocup) malware, which displays advertisements and installs additional malware. Stegoloader operators may install Vundo on a compromised system for additional monetary profit after they have extracted all the information they deem interesting.
Stegoloader is stealthy in many aspects; it evades analysis tools and deploys only necessary modules, without writing them to disk. There are likely more Stegoloader modules than CTU researchers have observed, possibly used by threat actors to ensure persistence or to gain access to additional resources. Although CTU researchers have not observed Stegoloader being used in targeted attacks, it has significant information stealing capabilities. Stegoloader is the third malware family that CTU researchers have observed using digital steganography. This technique might be a new trend because malware authors need to adapt to improved detection mechanisms.
The threat indicators in Table 5 can be used to detect activity related to Stegoloader.
|723ef64c6a1b1872bc84a9dc30e10c9199f5a153||SHA1 hash||Stegoloader deployment module executable|
|a48594b243f801e02066b77e46135382e890daf6||SHA1 hash||Stegoloader deployment module executable|
|c82c3d32211ea73b884cffe66cb1a46a080c5723||SHA1 hash||Stegoloader deployment module executable|
|68e3e19c14d2e10c67670999c77eb08221e16a08||SHA1 hash||Stegoloader deployment module executable|
|f6bb47621183060c2cd9df5a52face6eb1d52983||SHA1 hash||Stegoloader deployment module executable|
|b55497e02d61f059fe23cd86083eddfb0f718cdc||SHA1 hash||Stegoloader deployment module executable|
|eee347e8942c1ddc603e8c1a89dacf39673c2689||SHA1 hash||Stegoloader deployment module executable|
|5e1077fc19410b1dee59c11fd9cd7810c95ebaec||SHA1 hash||Stegoloader deployment module executable|
|d5d0a9ecf1601e9e50eef6b2ad25c57b56419cd1||SHA1 hash||Stegoloader deployment module executable|
|b8db99cf9c646bad027b34a66bb74b8b0bee295a||SHA1 hash||Stegoloader deployment module executable|
|3ad4376043d1297773e808a539ec0bd2f22b200c||SHA1 hash||Stegoloader deployment module executable|
|ccca1fbfdb1efaee8b6785879a4210a56e3e0d47||SHA1 hash||Stegoloader deployment module executable|
|43e1bfd48ee72d829c17ca1e8c9ecf296830ca8a||SHA1 hash||Stegoloader deployment module executable|
|Avanquest_PowerDesk_9_0_1_10_keygen.exe||Filename||Stegoloader deployment module|
|AVS_Video_Converter_9_1_1_568_keygen.exe||Filename||Stegoloader deployment module|
Table 5. Threat indicators for Stegoloader.
Dell SecureWorks. "Malware Analysis of the Lurk Downloader." April 23, 2014. https://www.secureworks.com/research/malware-analysis-of-the-lurk-downloader
Hex-Rays. "About the IDA Interactive Disassembler." https://www.hex-rays.com/products/ida/
Kafeine. "Inside Pony 1.7 / Fareit." June 26, 2012. http://malware.dontneedcoffee.com/2012/06/inside-pony-17.html
Microsoft. "Constant Special Item ID list (CSIDL)." https://msdn.microsoft.com/en-us/library/windows/desktop/bb762494(v=vs.85).aspx
Trend Micro. "TSPY_GATAK.GTK." http://www.trendmicro.com/vinfo/us/threat-encyclopedia/malware/tspy_gatak.gtk
Table 6. Log messages observed by CTU researchers.
|ProcessorArchitecture||Architecture of the processor (32-bits or 64-bits)|
|CountryName||Operating system's configured country|
|OwnerName||Name of the compromised computer's owner|
|CompanyName||Name of the compromised computer's company|
|UserName||Username of the logged-in user|
|UserRights||Rights of the logged-in user|
|UserRole||Role of the logged-in user|
|OsName||Name of the operating system (e.g., Microsoft Windows XP)|
|OsID||Operating system ID|
|OsSN||Operating system serial number|
|TimeZone||Time zone set in the operating system|
|InternalIP||IP address of local interface|
|PlatformVendor||Vendor for the CPU (e.g., VMWare Virtual Platform)|
|PlatformName||Name of the platform (e.g., VMWare Inc.)|
|ScreenResolution||Resolution of the main display|
|VideoCardVendor||Vendor name of the video card|
|VideoCardName||Name of the video card|
|Processes||List of processes currently running on the compromised system|
|SessionTime||Time elapsed since the current user logged in|
|RouterMAC||MAC address of the default gateway|
Table 7. Element names and descriptions of system data collected by Stegoloader.