Description

Researchers at K7 Labs have uncovered a highly sophisticated Python-based malware that uses advanced obfuscation and stealth techniques to inject malicious code into a legitimate Microsoft executable (cvtres.exe). The malware’s delivery chain begins with a PE dropper, which decrypts and reconstructs a batch script at runtime; this script establishes a working directory (C:\DragonAntivirus) and then downloads a file masquerading as a harmless .png image. Unbeknownst to defenders, this “image” is actually a RAR archive packed with malicious components. Once extracted via a bundled WinRAR executable, the archive yields three key files: a password-protected .sys file (actually a Python runtime), a JSON file that gets renamed to Install.bat, and a legitimate WinRAR helper binary. The batch script then launches the fake Python runtime (masquerading as ntoskrnl.exe) with specific arguments. These arguments trigger a multistage de-obfuscation pipeline: the payload is successively decoded via Base64, decompressed via BZ2, then zlib, before finally being reconstructed using Python’s marshal.loads to obtain executable bytecode. Crucially, only a small portion of the resulting blob is valid Python bytecode; much of it is filler used to hide the malicious content. After reconstruction, the malware injects itself into the cvtres.exe process a signed Microsoft binary and then downloads a .NET component from its C2 server, which establishes encrypted TCP-based communication for remote control. This strategy gives the attacker strong stealth and persistence: by hiding within a trusted system process, the malware can evade many traditional security controls while maintaining a persistent, encrypted remote-access channel. Malware Analysis, News and Indicators From an analysis perspective, this campaign reveals how threat actors are increasingly combining scripting languages like Python with native binaries and signed executables a blend that frustrates signature-based detection. The use of layered encoding and decoy files (like the faux .png) further complicates forensic detection and reverse engineering.