StealC Malware Hidden in Fake Qwen Model Repository
A GitHub repository impersonating Qwen distributed a 487 KB ZIP file advertised as a 27B-parameter local AI model. Instead of the expected model weights, the archive contained three files: a Windows launcher, a renamed LuaJIT interpreter and an obfuscated Lua script.
Security researchers from SlowMist’s MistEye system found that the script collected hostnames, usernames, system identifiers, operating-system details and screenshots. It sent the data to attacker-controlled infrastructure and could download and execute additional payloads. A Polygon smart contract was also used to provide a fallback command-and-control address, making infrastructure blocking more difficult.
Further analysis identified a second-stage StealC information-stealing Trojan. The malware targets Chrome, Edge, Brave and Firefox data, including passwords, cookies and browsing history. It can also collect Outlook, Foxmail, WinSCP and Steam credentials, screenshots, system information and potentially cryptocurrency wallet data controlled by server-side tasks. Its capabilities include persistence, secondary downloads, privilege escalation and encrypted data exfiltration.
MistEye identified 29 similar ZIP files across 23 repositories. Researchers stressed that the official Qwen project was not compromised and that the analysis was based on static evidence and simulated requests; samples were not executed on real hosts.
For crypto traders, the StealC malware poses a direct account and wallet-security risk. Users should avoid unofficial model repositories, verify file sizes and sources, and rotate credentials or wallet access if suspicious files were opened.
Bearish
The expected market impact is bearish, but likely limited at the broad market level. The incident does not indicate a compromise of Qwen, a major blockchain network or a cryptocurrency exchange. It is instead a targeted malware campaign using a fake AI model to reach users who may hold exchange credentials, browser sessions or wallet data.
The short-term effect could include increased risk aversion among crypto users, warnings from security firms and isolated wallet thefts. If stolen cookies, passwords or private keys are used, affected accounts could see unauthorized withdrawals or token selling. Such incidents can create temporary selling pressure in the assets linked to compromised wallets, although the article provides no evidence of a systemic market liquidation.
The use of Polygon infrastructure for fallback command and control does not imply a flaw in the Polygon network or POL. However, it highlights how attackers can use public blockchains to rotate infrastructure, potentially extending campaigns and complicating incident response. Similar malware outbreaks and supply-chain attacks have historically produced short-lived, asset-specific volatility rather than sustained market-wide declines.
Over the longer term, repeated wallet-stealing campaigns could damage user confidence in unofficial software, increase demand for hardware wallets and transaction screening, and encourage exchanges to strengthen session protection. Traders should monitor wallet outflows, exchange security alerts and on-chain movements linked to known StealC indicators. The primary risk is operational and reputational, not a fundamental change to crypto valuations.