Overview

SonicWall Capture Labs threat research team has observed an updated variant of StrelaStealer. StrelaStealer is an infostealer malware known for targeting Spanish-speaking users and focuses on stealing email account credentials from Outlook and Thunderbird. StrelaStealer was reported in the wild in early November 2022. StrelaStealer has been updated with an obfuscation technique and anti-analysis technique.

Technical Analysis

MD5: 1E37C3902284DD865C20220A9EF8B6A9

SHA256: F2D7CF39392D394D6CCD0F9372DB7D486D4CB2BB6C3BBFD0D8BFBB6117A5E211

This updated version of malware delivered via JavaScript comes in archive files as attachments in emails. The initial vector is JavaScript which will drop the 64-bit executable file in the %userprofile% folder and execute the malware process. We have observed that StrelaStealer is being delivered as a 64-bit exe as well as a DLL via JavaScript. We are explaining the analysis for the 64-bit executable in this blog. This 64-bit executable is a wrapper that will act as a loader for the actual payload.

In the main 64-bit executable file, the data section has an encryption key, and the size of the encryption key is 0x2714 bytes. The encoded payload is embedded in the data section at the end of the encryption key. The size of the payload is 0x1C600. A single-byte XOR encryption is performed to decrypt an encoded PE file from the data section.

Figure 1:  Encryption key started from 0x10th offset in the data section

Figure 2:  Obfuscated Jumps

Figure 3: Graph view for obfuscated function

Figure 4: Another graph view of the obfuscated function

Figure 5:  PEB parsing code fragments inside the jump code block

This obfuscation is quite effective. Anti-analysis techniques delay the execution, and the researcher has to search the code fragments inside the jump blocks, which is a tedious task.

Along with jump blocks and multiple loops, there are multiple dummy functions that are not doing anything but wasting time while analyzing the sample.

Figure 6: Dummy functions inside nested Jumps

Figure 7: Dummy functions

Figure 8: XOR decryption to decrypt the encoded payload

Once it decrypts the payload, it reads the encoded API string array at the end of the encoded payload embedded in the data section. Within the payload, the first DWORD is the size of the array and next is the API function array. This array is of size 0x52 bytes and the encryption key used earlier to decrypt the payload will also be used to decrypt the API array. The only difference between the decryption of the payload and the array  is malware uses an encryption key of size 0x52 bytes from the 4^th offset of encryption key.

Figure 9: Encoded API array

Figure 10: Malware calculates the start offset of the encoded API string and starts decrypting it

Figure 11: API array after an XOR decryption

It accesses the PEB structure and parses it to get the list of loaded modules in process memory.

The following is an example of the instructions set to parse the PEB.

Figure 11B: Instructions

Here InLoadOrderModuleList is a doubly-linked list that contains the loaded modules for the process.

The malware parses this “InLoadOrderModuleList” to get the Imagebase address of kernel32.dll with the goal of resolving the VirutalAlloc API Then the malware will parse the PE structure of kernel32.dll to get the name of each exported function and matches them with the API string that got decrypted earlier in 0x52 byte array. If the API name matches the exported function name, then the malware will read the associated function RVA from the export directory and add it to the Imagebase of kernel32.dll,. Using this method, the malware resolves each API dynamically. It will resolve 4 APIs – here VirtualAlloc,  LoadLibraryA , GetProcAddress , and MessageBoxTimeoutA. Once its finished resolving the APIs, the malware will show the error massage box and then continue execution.

Now, the malware calls the “VirtualAlloc” API to allocate memory in the process and start its task as loader to load the actual payload.

• The malware parses the PE file structure of the payload from the data section where previously it decrypted the PE file and read each section header one by one.
• To map the process as per section alignment, it reads the virtual address of each section and adds it to the image base of the injected PE and copies each section of data to this offset in memory.
• The malware will not copy the PE header to the injected PE, this has been done intentionally to evade detection from AV products.
• It reads the relocation section and does the fixup as it gets loaded at the different base address in the memory.
• It reads the import address table of the payload file from the data section region and resolves the API address dynamically using the “LoadLibraryA” and “GetProcAddress”  APIs and copies these all function pointers to the IAT of the injected payload.
• When the injected PE file is ready for execution, it will read the RVA of the address from the entry point from the PE file in the data section and add the base address of the injected payload and redirect execution to the injected code.

Figure 12: Configuration setting for the payload

The injected payload is 64-bit executable file, it will call the “GetKeyboardLayout” API and check the lower words of the return value with the hardcoded values in binary. It tries to check if the keyboard layout is from the following countries. If it is, then the malware will continue its execution, otherwise it terminates itself.

Figure 13: Call to the “GetKeyboardLayout” API and check language identifiers

Now, the payload retrieves the computer name by calling the “GetComputerNameA” API and encrypts the first 4 bytes of the computer name string using single byte XOR encryption. The encryption key is “MIR24”, which is hardcoded in binary. It will create a Mutex with the name of this partially encrypted computer name string. If a Mutex already exists, it will terminate it.

Figure 14: Creating a Mutex and executing its core functionality to steal data from the infected machine

As we can see in Figure 14, it will execute the function which will steal confidential data from the infected machine.

Here, we have found two functions in the malware. The first is used to steal data from Mozilla Thunderbird, which is a free and open-source email client software. The other function is intended to steal data from Outlook.

• It searches for the folder path “C:\Users\<username>\AppData\Roaming\Thunderbird\Profiles\”

All of your data such as messages, passwords and user preferences as well as changes made while you use Thunderbird are stored in a special folder called profile.

• If it finds this folder path on the system, it will call the FindFirstFileAand FindNextFileA APIs to search for two files in the subdirectory. The first is “logins.json” (account and password) and the second is “key4.db” (password database).
• It reads the data from both of these files and appends both files’ data one after another, starting network communication.
• It establishes a connection to its server and prepares an HTTP post request with the user-agent “Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/60.0.3112.113 Safari/537.36” and then exfiltrates this data to its server.

http[:]//45[.]9[.]74.12/server.php .

• The server IP is hardcoded in binary which is “45.9.74[.]12”
• Before sending data to the server, it will encrypt it with the single byte XOR encryption. The encryption key is hardcoded in binary which is “00ca8abe-6ab2-4b10-97c8-925934cf0423”

Figure 15: Searches for the “logins.json” and “key4.db” files from the profile folder

Figure 16: StrelaStealer is expecting the response from its server

We have analysed the second function statically where it reads the windows registry key, enumerates data from it and tries to locate the ‘IMAP User’, ‘IMAP Server’ and ‘IMAP Password’ values.

The IMAP Password contains the user password in encrypted form. The malware will call the Windows “CryptUnprotectData”  API to decrypt it.

The following registry key is enumerated to steal Outlook data:

“SOFTWARE\Microsoft\Office\16.0\Outlook\Profiles\Outlook\9375CFF0413111d3B88A00104B2A6676\”

Figure 17: Outlook registry key would have been enumerated to steal data from the infected machine

Figure 18: Network communication with server

The archive file cannot be found in any of the popular threat intelligence sharing portals like VirusTotal at the time of writing this blog.

Figure 19: File is not available on VirusTotal

This threat is detected by SonicWall Capture ATP w/RTDMI . Evidence of the detection by our RTDMI engine can be seen below in the Capture ATP report for this file.

Figure 20: Capture report

IOCs

Archive file
MD5: ca4797bf995c91864c8b290ebd4e1c7b
SHA256: 74f21472fed71aaccbd60b34615a8390725cbab6cb25bbc6a51bd723ff8bd01a

JavaScript (Initial vector)
Md5 : C235CE3765F9B1606BDA81E96B71C23B
SHA256 : E083662C896C47064FD47411D47459BF4B1CB26847B5D26AEDD7F9D701CABD43

Main 64-bit executable file
MD5 : 1E37C3902284DD865C20220A9EF8B6A9
SHA256 : F2D7CF39392D394D6CCD0F9372DB7D486D4CB2BB6C3BBFD0D8BFBB6117A5E211

Injected 64-bit Payload
MD5 : 95F51B48FB079ED4E5F3499D45B7F14E
SHA256 : C02BB26582576261645271763A17DE925C2D90D430E723204BAEC82030DC889A

Server IP : “45[.]9.74[.]12”

Overview

The SonicWall Capture Labs threat research team became aware of a noteworthy vulnerability — an Unauthenticated Command Injection — in Progress Kemp Loadmaster, assessed its impact and developed mitigation measures for it. Kemp Technologies’ LoadMaster, an application delivery controller and load balancer, published an advisory on this vulnerability affecting all LoadMaster releases after 7.2.48.1 and the LoadMaster Multi-Tenant (MT) VFNs. LoadMaster can be deployed on various platforms such as hardware, cloud and virtual machines. This vulnerability is identified as CVE-2024-1212 and was assigned a critical CVSS score of 9.8.  Considering the sizeable user base, low attack complexity and publicly available exploit code including a Metasploit module, LoadMaster users are strongly encouraged to upgrade their instances to the latest versions with utmost priority.

Technical Overview

This vulnerability allows threat actors to circumvent the authentication mechanism by sending a crafted request to the web server.

The conditions that led to the exploitation of the vulnerability in the Progress Kemp LoadMaster load balancer were:

• Bypassing Disabled API Restrictions: It was possible to access the REST API, even when disabled, by crafting a specific request path and parameters. This bypass allowed researchers to reach critical functions that were supposed to be inaccessible with the API disabled.
• Unauthenticated User Input Handling: The system did not properly validate or sanitize the “REMOTE_USER” and “REMOTE_PASS” environment variables, which were set based on user-provided basic authentication headers. This flaw allowed for the injection of arbitrary commands.
• Command Injection via System Call: The lack of validation and sanitization led to constructing a command with user-controllable input that was then passed to a system() call. This behavior facilitated the execution of arbitrary commands on the system.
• Exploiting Basic Authentication for Command Execution: By manipulating the base64-encoded authorization string sent in the HTTP headers, attackers could inject commands that the server would execute, enabling a direct path to command injection and system compromise.

Triggering the Vulnerability

The flaw is in the processing of the “/access/” RESTful API Interface to the LoadMaster.

Figure 1: LoadMaster CGI Bash Script

As seen in the code shared by RhinoSecurity Labs in Figure 1,  user input to the “/access/” API is directly put into a bash script leading to a critical vulnerability tracked as CVE-2024-1212. Before being passed to any function(s), the user input should be mandatorily parameterized and sanitized.

Leveraging this unauthenticated command injection vulnerability requires access to the vulnerable LoadMaster administrator web user interface. The publicly available also shows the possibility of privilege escalation once the shell is obtained.

An example request to trigger the vulnerability would look like this http[:]//target-ip:port/access/set?param=enableapi&value=1 with the Authorization parameter containing the command injection, as shown in Figure 2.

Figure 2: Triggering CVE-2024-1212 PoC packet capture

Notice the command injection is base64 encoded. When decoded, the attacker is sending the ‘;echo ‘[S]’hostname;echo’[E]’;’:anything as shown in Figure 3.

Figure 3: Decoded-authorization-header

The default admin configuration of a LoadMaster instance is a user named “bal”, as shown in Figure 4. One can fully control the system by manipulating sudo user entries via the management interface.

Figure 4: Default LoadMaster interface

SonicWall Protections

To ensure SonicWall customers are prepared for any exploitation that may occur due to this vulnerability, the following signatures have been released:

• IPS 4362 – Progress Kemp LoadMaster Command Injection

Threat Graphs

SonicWall sensors have confirmed exploitation attempts of this vulnerability. The graph below indicates increased exploitation attempts over the last 6 days.

Figure 5: Threat graph

Remediation Recommendations

Considering the severe consequences of this vulnerability and the trending of unauthenticated nefarious activists trying to get Loadmaster management interface access using the exploit in the wild, users are strongly encouraged to upgrade their instances as published in the vendor advisory.

Relevant Links

• NVD
• GitHub PoC
• Latest LoadMaster-Firmware
• GitHub Advisory Database
• RhinoSecurity Labs blog
• LoadMaster Advisory
• LoadMaster Release Notice
• LoadMaster ECS Connection Manager Advisory

*Originally this article claimed we had confirmed active exploitation of this vulnerability. This was a mistake. We have confirmed exploitation attempts, and the article now reflects that.

Overview

This week, the Sonicwall Capture Labs threat research team analyzed a new Golang malware sample. It uses multiple geographic checks and publicly available packages to screenshot the system before installing a root certificate to the Windows registry for HTTPS communications to the C2. There is currently no malware family affiliated, but the IP and URL addresses have been used by AgentTesla, GuLoader, PureLog Stealer and others.

Technical Analysis

The sample is detected as a Golang 64-bit executable with a WinAuth certificate. The timestamp has been tampered with, as it shows a creation date of December 31, 1969.

Figure 1: Sample detection with Detect It Easy

The certificate listed is issued from ‘RUNNINGWITHSCISSORS LTD’ but does not have an issuer signatory.

Figure 2: There is no signer, email or creation date with this certificate

Looking at the sample, one section included is non-standard, .symtab; this section is associated with ELF/Linux files and should not be on a Windows executable.

Figure 3: This is a Linux file-type section and would not be accessed by normal means

Because of the way Golang compiles binaries, the program had to have the gopcln table reconstructed in order to see the inner workings of the file.

Figure 4: Malware functions in cleartext

Once done, there are several items of note. Newly renamed functions list out the primary methods of . The strings show a pattern that looks to be used to send system information to the attacker – this is confirmed later. This pattern contains information for a system UUID, remote IP, username, hostname, Windows version, process ID, process name and architecture.

Figure 5: There is an ASCII pattern for information that may be sent to the attacker

Next, the packages reveal that the author is using public GitHub applications from the following repositories to generate and save screenshots:

• https://github.com/kbinani/screenshot
• https://github.com/lxn/win
• https://github.com/fogleman/gg
• https://github.com/golang/freetype

Figure 6: These GitHub packages are public and are not malicious on their own

During runtime, the malware will start by querying main system information using process injection via VirtualAlloc and running WMIC. The following two commands are run first:

• ‘SELECT UUID FROM Win32_ComputerSystemProduct’
• C:\Windows\System32\conhost.exe C:\Windows\system32\conhost.exe 0xffffffff -ForceV1

Next, these registry keys are queried for the Windows version and hostname:

• HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion
• HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Services\Tcpip\Parameters

When attempting to control/open WMIC, the sample looked in the following locations, as well as trying to find a file with an unusual name:

• C:\Users\user\Desktop
• C:\Program Files (x86)\Common Files\Oracle\Java\javapath\
• C:\Windows
  C:\Windows\system32
• C:\Windows\Wbem\

Figure 7: Sample of locations WMIC is searched for along with an unusual file

Once WMIC has been launched, both the main executable and WMIC will load, or attempt to load, the following DLLs:

Apphelp.dll
Cryptbase.dll
Winmm.dll
Powrprof.dll
Umpdc.dll
Userenv.dll
Profapi.dll
Netapi32.dll
Wkscli.dll
Netutils.dll
Samcli.dll
Samlib.dll
Iphlpapi.dll
Dhcpcsvc6.dll
Dhcpcsvc.dll
Dnsapi.dll
Mswsock.dll
Rasadhlp.dll
Fwpuclnt.dll
Msasn1.dll
Cryptsp.dll
Rsaenh.dll
Gpapi.dll
Framedynos.dll
Sspicli.dll
Kernel.appcore.dll
Wbemcomm.dll
Msxml6.dll
Urlmon.dll
Iertutil.dll
Srvcli.dll
Uxtheme.dll
Vcruntime140.dll
Vcruntime140_1.dll
Amsi.dll
Vbscript.dll
sxs.dll

The majority of these were not used during testing, which means that they may be used after initial contact is made with the C2 for further data collection or additional malware utilities.

Figure 8: Sample listing of DLLs searched for by the trojan

When the sample reaches out on the network, it attempts to get an IP using ‘ip-api.com/json’ and to make initial contact with the C2 at ‘https://daily-mashriq[.]org/goyxdrkhjilchyigflztv’ using a ping.

Once those steps are complete, the malware installs a new root certificate by overriding data in the following registry key using ‘CertGetCertificateChain’:

HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\SystemCertificates\ROOT\Certificates\3F728A35DE52B2C8994A4FB101A03B95E87B06C8

Figure 9: Previous data being overwritten by the new certificate

Although the main functions go into detail about screenshots, there are also API calls in memory for tracking other Windows events. These are monitored with SetWindowsHookExW and TrackMouseEvent. Screenshots are created using the GitHub packages previously mentioned.

Figures 10, 11: Code showing that the ‘kbinani’ takes the screenshot, then ‘fogleman’ saves it.

Network connection to ‘daily-mashriq[.]org’ is attempted with the previously discovered pattern containing information encoded with chacha20 and sent using User-Agent “AGCYRNRWWWFZZSWWFWDYDCVDN”:

Figure 12: Encoded information sent to the C2

However, after the initial message has attempted to POST, subsequent messages only include the UID and are sent every seven seconds.

Figure 13: Repeated POSTS only contain the UID

The following ports are also bound for listening: 49708-49711, 49720-49730, and 49733-49750.

As of this writing, it appears that the domain has been blocked by (CHECK THIS).

SonicWall Protections

SonicWall Capture Labs provides protection against this threat via the following signature:

• MalAgent.Go.Trojan

IOCs

• 8f4cf379ee2bef6b60fec792d36895dce3929bf26d0533fbb1fdb41988df7301
• https://daily-mashriq.org/goyxdrkhjilchyigflztv
• http://ip-api.com/json/