Info Stealing Packages Hidden in PyPI

Affected platforms: All platforms where PyPI packages can be installed
Impacted parties: Any individuals or institutions that have these malicious packages installed
Impact: Leak of credentials, sensitive information, etc.
Severity level: High

The Python Package Index (PyPI) is an open repository of software packages developed by the Python community to help people quickly develop or update applications. While most of the packages uploaded to PyPI are posted by dedicated individuals looking to support the Python community, threat actors also regularly post packages infected with malware. The FortiGuard Labs team uses a proprietary, AI-driven OSS malware detection system to hunt for and monitor these threats. Recently, we identified a PyPI malware author (who goes by the ID “WS”) discreetly uploading malicious packages to PyPI. We now estimate that there may be well over 2000 victims of “WS” just from the packages described below alone.

The identified packages—nigpal, figflix, telerer, seGMM, fbdebug, sGMM, myGens, NewGends, and TestLibs111—exhibit attack methodologies similar to those outlined in a Checkmarx blog post published four months ago. The similarity suggests a possible connection to a malicious campaign from early 2023. Notably, these packages incorporate base64-encoded source code of PE or other Python scripts within their setup.py files. Depending on the victim devices’ operating system, the final malicious payload is dropped and executed when these Python packages are installed.

A brief timeline of our findings about this info-stealing author is detailed below.

Figure 1: A brief timeline of malicious PyPI packages published by the author “WS”

The packages released before December 2023 are very similar to those discussed in earlier blog posts. Specifically, they deploy Whitesnake PE malware if the victim’s device runs on Windows, or they can deliver a Python script designed to steal information from Linux devices. A subtle distinction lies in the new method now being used by the Python script to transmit stolen data. Instead of relying on a single fixed URL, these new malware variants use a range of IP addresses as the destination, likely to ensure successful data transmission even if one server fails. Given the strong resemblance of these new packages to the previous ones, we will focus on analyzing the payloads from these more recent variants.

Unlike earlier attacks that targeted both Windows and Linux users, this recent set of packages predominantly targets Windows users. While the executable payload for each package varies slightly from one another, they consistently aim to exfiltrate sensitive information from victims.

Let’s take a brief look at a few of these payloads.

PE Payload of sGMM Package

Unlike the previous Whitesnake PE payloads developed using .NET, this sample manifests as a Python-compiled executable crafted using the PyInstaller tool.

Upon careful examination of its contents post-extraction, the primary script file is revealed as ‘main.pyc’ accompanied by an ‘addresses.py’ counterpart. The mere presence of this counterpart raises suspicions.

Figure 2: Extracted files

Upon decompiling ‘main.pyc,’ we can see that the code is somewhat incomplete, accentuating its clandestine nature. Red flags include copying itself to the Windows startup folder for autorun on startup. The script also probes logical drives, endeavoring to copy itself into non-removable drives. Notably, it adopts a discreet strategy by monitoring the count of its running instances, ensuring an inconspicuous profile complete with an exit strategy if the count surpasses two.

Figure 3: Decompiled main.pyc

Finally, the script retrieves the clipboard contents and compares them against predefined cryptocurrency address patterns. The discovery of matching patterns prompts the script to overwrite the clipboard with corresponding addresses sourced from ‘addresses.py.’ This could deceive unsuspecting victims into directing things like cryptocurrency transactions to an unexpected destination.

Figure 4: addresses.py

PE Payload of myGens & NewGends Packages

This payload is an encrypted .NET executable. Upon being installed on the user’s device, the payload launches an invisible window of “cmd.exe.” It then uses “powershell.exe” through this window to add itself to the Windows Defender’s exclusion list to bypass the built-in security response:

Figure 5: Adding itself to the Windows Defender exclusion list

It then copies itself to the Windows Application folder and creates a scheduled task to run itself every hour once the device is infected:

• %LOCALAPPDATA%\Packages\Microsoft.Windows.Accounts.ControlRCP_ruzxpnew4af
• HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Schedule\TaskCache\Tasks\{070B9798-7C59-4A69-BB25-F353628999B0}\Path|\MicrosoftWindowsControlRCP

Figure 6: Confirmation of copying itself and scheduling a task

Each time the task runs, it establishes a connection to the malicious IP: 194[.]36[.]177[.]30. It uses “socket.io” to connect the server to the client. This way, if the connection is lost, it will automatically try to reconnect. 

Figure 7: Decryption of “socket.io” and IP

The core of this payload tries to collect user information, such as the user’s IP address, the host credentials, etc.:

Figure 8: User info grabber

The following strings offer more compelling evidence of the information-stealing capabilities of this payload. They reveal that the payload not only captures mouse and keyboard interactions but also acquires and transmits wallet and browser data to the remote server.

Figure 9: List of Properties for the stealer

Figure 10: Snippet of code showing wallet grabbing

Figure 11: Function showing browser stealing

PE payload of TestLibs111 Package

Once this sample runs, we can observe data being sent to a suspicious IP address (194[.]36[.]177[.]30), as shown in Figure 12. At first glance, we might dismiss these as inconspicuous strings. However, further investigation shows this to be a .zip file shrouded in multiple layers of encryption. This raises serious concerns about its behavior.

Figure 12: Traffic sent to IP address 194[.]36[.]177[.]30

Upon decryption, we see ‘PK’ as the initial bytes, indicating a .zip file format.

Figure 13: Hex dump of .zip file

After debugging this sample, there are several indicators as to what sensitive content it may be interested in. Below are a few of the many interesting strings that were found during our analysis.

Figure 14: Strings found in assembly code

Some of these strings may indicate info-stealing from different services, such as browsers, applications, and cryptocurrency services. Some examples we found included Chrome, Discord, Filezilla, and Coinbase.

When we extract the .zip file, we can see some of the collected content.

Figure 15: zip file contents

The author then exfiltrates this sensitive information to the server.

Conclusion

This blog demonstrates the ability of a single malware author to disseminate numerous info-stealing malware packages into the PyPI library over time, each featuring distinct payload intricacies. Users are urged to exercise utmost caution when using open-source packages, checking for malicious content or payloads that may render targeted devices susceptible to information theft. The narrative of this particular malware author has unfolded over several months and showcases the considerable amount of havoc that has been wrought. Information-stealing malware is an increasingly pertinent and pressing subject. Safeguarding against such persistent adversaries demands a strategic and forward-thinking approach to fortify your defenses.

Fortinet Protections

FortiGuard AntiVirus detects the malicious files identified in this report as

nigpal-0.1 setup.py: PYTHON/Agent.c51b!tr
nigpal payload: (PE) MSIL/WhiteSnake.C!tr
nigpal payload: (Python) PYTHON/Agent.7f73!tr
figflix-0.1 setup.py: PYTHON/Agent.c51b!tr
figflix-0.2 setup.py: PYTHON/Agent.c51b!tr
figflix payload: (PE) MSIL/WhiteSnake.C!tr
figflix payload: (Python) PYTHON/Agent.7f73!tr
telerer-2.0 setup.py: PYTHON/Agent.c51b!tr
telerer-2.1 setup.py: PYTHON/Agent.c51b!tr
telerer-2.2 setup.py: PYTHON/Agent.c51b!tr
telerer-2.3 setup.py: PYTHON/Agent.c51b!tr
telerer payload: W32/Kryptik.HTVT!tr
seGMM-1.3.1 setup.py: PYTHON/Agent.c51b!tr
seGMM-3.1.1 setup.py: PYTHON/Agent.c51b!tr
seGMM payload: W32/Kryptik.HTVT!tr
fbdebug-0.1 setup.py: PYTHON/Agent.c51b!tr
fbdebug payload (PE): MSIL/WhiteSnake.C!tr
fbdebug payload (Python): PYTHON/Agent.7f73!tr
sGMM-1.3.2 setup.py: PYTHON/Agent.c51b!tr
sGMM payload: W32/ClipBanker.DE!tr
myGens-1.3 setup.py: PYTHON/Agent.c51b!tr
myGens-1.3 payload: MSIL/Agent.ERW!tr.spy
NewGends-1.3 setup.py: PYTHON/Agent.c51b!tr
NewGends-1.3 payload: MSIL/Agent.ERW!tr.spy
TestLibs111-1.4 setup.py: PYTHON/Agent.c51b!tr
TestLibs111-1.5 setup.py: PYTHON/Agent.c51b!tr
TestLibs111-1.7 setup.py: PYTHON/Agent.c51b!tr
TestLibs111 payload: W32/Agent.ORR!tr.pws

The FortiGuard AntiVirus service is supported by FortiGate, FortiMail, FortiClient, and FortiEDR. Customers running current AntiVirus updates are protected.

The FortiGuard Web Filtering Service detects and blocks the download URLs cited in this report as Malicious.

The FortiDevSec SCA scanner detects malicious packages, including those cited in this report that may operate as dependencies in users’ projects in test phases, and prevents those dependencies from being introduced into users’ products.

If you believe these or any other cybersecurity threat has impacted your organization, please contact our Global FortiGuard Incident Response Team.

IOCs

File	Hash (sha256)	Detection
nigpal-0.1 setup.py	c53d1387864ea3034bc4e19af492b3e67147d3fdc1d8b9752e24600d6919e3af	PYTHON/Agent.c51b!tr
nigpal payload (PE)	ab75ea75d1fe5bc51ecef274a95f7b835b09a0c7c95c4227366a3d64b5dee7c0	MSIL/WhiteSnake.C!tr
nigpal payload (Python)	24e07dd8c4a6fb92d842ebc168a40505bbd0421a16c13a06571910ca7a40a5a5	PYTHON/Agent.7f73!tr
figflix-0.1 setup.py	41ff3fedb78c672c6d0e5e849f81c8be10c0767558fcfdf6f529215556354d9e	PYTHON/Agent.c51b!tr
figflix-0.2 setup.py	377e8ca04aed57a10b350d9eb4a6e64818bb69b790f33db4be0fd22589c435ad	PYTHON/Agent.c51b!tr
figflix payload (PE)	34e5bd67fbd9a7040dca9cae90e36013aaeda1940bb39e7fcd5d5fa9c85cadc8	MSIL/WhiteSnake.C!tr
figflix payload (Python)	b2bf755c4e1336f5ab36bc679d4a86e4c0d4da7b33a26b9ec8c01e179027f66b	PYTHON/Agent.7f73!tr
telerer-2.0 setup.py	3dcff80475ebfb9a3aa93f3cebd8f008ea64d857a7c53719f1ca047dfd050e1c	PYTHON/Agent.c51b!tr
telerer-2.1 setup.py	dc5b74c1007bbe9acce3cddf30870766867b40e7d37264b7bdaf3b5f40747c10	PYTHON/Agent.c51b!tr
telerer-2.2 setup.py	f1f6501a97b9145d8dd755d25a39c1803fe54995a39fd59b2914f591d56bdc68	PYTHON/Agent.c51b!tr
telerer-2.3 setup.py	8bdc674e3a41370a2d0a997b6ca673c6d646ed580400af242980a5ec374864a2	PYTHON/Agent.c51b!tr
telerer payload	4fac457f8170e26643d0a4d8a0199e93d72872e1799e95f5c522a50754982079	W32/Kryptik.HTVT!tr
seGMM-1.3.1 setup.py	ef0e1a8378d1dd9e3cfd0d59d1969b618e15ddb4bbfaf50057670842004346e8	PYTHON/Agent.c51b!tr
seGMM-3.1.1 setup.py	6fe87ab0590229d11f2d174bcf13cfbaca6f6c9dc55af84527c96de16c12c799	PYTHON/Agent.c51b!tr
seGMM payload	4fac457f8170e26643d0a4d8a0199e93d72872e1799e95f5c522a50754982079	W32/Kryptik.HTVT!tr
fbdebug-0.1 setup.py	da0c21c66fd0dc42b1bcb06c9bd0d7e48b1b866720229712df64410eebd62199	PYTHON/Agent.c51b!tr
fbdebug payload (PE)	d9568da21005794d80eb6572ccce47cc766ba5fe24b2b82cd4ff2cf05d8531a2	MSIL/WhiteSnake.C!tr
fbdebug payload (Python)	ec9e1342b0bddbd0ef65cd37a751b3a8c3c8170cdd8cce0f0fb6815b6be26a45	PYTHON/Agent.7f73!tr
sGMM-1.3.2 setup.py	8fb72c3a6a5d96f91c3dc46541331ebf0a6cf326d2353ab6f2b1c119e9907670	PYTHON/Agent.c51b!tr
sGMM payload	94be6da31c5f896017af733a44b9ea00abbb35bce0a8dbcab776367234e4d818	W32/ClipBanker.DE!tr
myGens-1.0 setup.py	f22110ea2376082651f5f0724875e6f9d083e2be0688dc06b59206c35fa50def	PYTHON/Agent.c51b!tr
myGens payload	03a1621af484ff8f5c1797b25426bab656b6731dba43e31fe58fc1f1963d8484	MSIL/Agent.ERW!tr.spy
NewGends-1.3 setup.py	857bc70fb5968b9f5e257e41f4be9cdd8c7135314bf6200e2cf5b60186401e7a	PYTHON/Agent.c51b!tr
NewGends payload	03a1621af484ff8f5c1797b25426bab656b6731dba43e31fe58fc1f1963d8484	MSIL/Agent.ERW!tr.spy
TestLibs111-1.4 setup.py	c9e0b8c6c5140acae2b3bf003d9ae2a69abf04253b0bd932ec97c732a4b9bf97	PYTHON/Agent.c51b!tr
TestLibs111-1.5 setup.py	0e13bb49aba0878b919bc0980ce2e9e3cfa876387fcedf5af41235ab0f7a440a	PYTHON/Agent.c51b!tr
TestLibs111-1.7 setup.py	14cd40cce030bfca6a4c06fdadd353b5eaa092e7f73ba65308afedc04270c9b9	PYTHON/Agent.c51b!tr
TestLibs111 payload	2b617277fc551b7500867ee009a0f80cbe6d5ee729bdfbf9b4f9d52164811082	W32/Agent.ORR!tr.pws

IP Address

194[.]36[.]177[.]30

95[.]140[.]147[.]126[:]8080

78[.]46[.]66[.]9[:]8080

116[.]203[.]194[.]247[:]8080

206[.]189[.]109[.]146[:]80

217[.]145[.]238[.]175[:]80

164[.]90[.]185[.]9[:]443

185[.]216[.]26[.]127[:]8080

94[.]156[.]6[.]209[:]80

185[.]217[.]98[.]121[:]443

185[.]217[.]98[.]121[:]80

192[.]99[.]44[.]107[:]8080

52[.]86[.]18[.]77[:]8080

18[.]218[.]18[.]183[:]80

192[.]99[.]196[.]191[:]443

216[.]250[.]190[.]139[:]80

24[.]199[.]110[.]250[:]8080

35[.]166[.]49[.]216[:]8080

5[.]78[.]68[.]6[:]8009

44[.]228[.]161[.]50[:]443

162[.]33[.]178[.]113[:]80

103[.]226[.]125[.]218[:]80

106[.]15[.]66[.]6[:]8080

121[.]63[.]250[.]132[:]88

18[.]228[.]80[.]130[:]80

65[.]20[.]76[.]112[:]80

47[.]96[.]78[.]224[:]8080

216[.]39[.]242[.]18[:]8080

129[.]151[.]109[.]160[:]8080

168[.]138[.]211[.]88[:]8099

139[.]99[.]123[.]53[:]9191

54[.]92[.]18[.]154[:]443

135[.]181[.]98[.]45[:]8888

65[.]108[.]226[.]108[:]8080

45[.]155[.]171[.]134[:]8080

141[.]94[.]175[.]31[:]8098

116[.]202[.]101[.]219[:]8080

129[.]159[.]134[.]19[:]8080

107[.]161[.]20[.]142[:]8080

3[.]142[.]76[.]113[:]80

104[.]184[.]140[.]41[:]9000

205[.]185[.]123[.]66[:]8080

154[.]31[.]165[.]232[:]80

139[.]84[.]231[.]199[:]8080

103[.]244[.]151[.]46[:]8080

52[.]196[.]241[.]27[:]443

13[.]112[.]250[.]213[:]443

189[.]115[.]63[.]77[:]8080