Weaponizing Trust Signals: Claude Code Lures and GitHub Release Payloads
Following an accidental leak of Anthropic's Claude Code source material, threat actors rapidly deployed a social engineering campaign using fake GitHub repositories. The campaign distributes trojanized archives containing a Rust-compiled dropper that deploys Vidar stealer and GhostSocks proxy malware, specifically targeting developers seeking AI tools.
Authors: Jacob Santos, Sophia Nilette Robles
Source:Trend Micro
- urlgithub[.]com/ai-wormGPT/wormGPT/releases/Malicious GitHub Release URL distributing trojanized AI tool archives.
- urlgithub[.]com/claude-ai-opus-4-6/claude-opus-4.6/releases/Malicious GitHub Release URL distributing trojanized AI tool archives.
- urlgithub[.]com/Kawaii-GPT-ai/KawaiiGPT/releases/Malicious GitHub Release URL distributing trojanized AI tool archives.
Key Takeaways
- Anthropic inadvertently exposed internal Claude Code source material via a misconfigured npm package, leaking 512,000 lines of TypeScript.
- Within 48 hours, threat actors leveraged the leak to distribute Vidar stealer and GhostSocks proxy malware via fake GitHub repositories.
- The campaign is part of a broader operation active since February 2026, impersonating over 25 software brands to deliver a Rust-compiled dropper.
- Attackers abuse GitHub Releases to deliver large trojanized 7z archives, using disposable accounts to evade takedowns.
- The leaked source code introduces long-term risks, including vulnerability discovery, prompt injection blueprinting, and agentic attack surface exposure.
Affected Systems
- Windows
- Developer Environments
Attack Chain
Victims search for trending software like Claude Code and land on malicious GitHub repositories. They download a trojanized 7z archive (78-167 MB) from GitHub Releases containing a Rust-compiled dropper, often named TradeAI.exe or ClaudeCode_x64.exe. Upon execution, the dropper deploys Vidar Stealer and GhostSocks proxy malware. Vidar uses Steam Community profiles and Telegram channels as dead drop resolvers to locate its C2 server, to which it exfiltrates stolen browser credentials, session tokens, and cryptocurrency wallets.
Detection Availability
- YARA Rules: No
- Sigma Rules: No
- Snort/Suricata Rules: No
- KQL Queries: No
- Splunk SPL Queries: No
- EQL Queries: No
- Other Detection Logic: No
The article does not provide specific detection rules (YARA, Sigma, etc.), but outlines behavioral indicators and file naming conventions for hunting.
Detection Engineering Assessment
EDR Visibility: High — EDR solutions can detect the execution of the Rust-compiled dropper, the subsequent deployment of Vidar and GhostSocks, and credential access behaviors targeting browsers. Network Visibility: Medium — Network monitoring can identify connections to GitHub Releases for large archive downloads, as well as traffic to Steam/Telegram for dead drop resolution and SOCKS5 proxy activity. Detection Difficulty: Moderate — Initial downloads from GitHub Releases blend with legitimate developer activity, but the subsequent execution of a Rust dropper and connections to dead drop resolvers provide solid behavioral detection opportunities.
Required Log Sources
- Process Creation (Event ID 4688 / Sysmon 1)
- Network Connections (Sysmon 3)
- File Creation (Sysmon 11)
Hunting Hypotheses
| Hypothesis | Telemetry | ATT&CK Stage | FP Risk |
|---|---|---|---|
| Look for unusual process execution originating from recently extracted large .7z archives downloaded from GitHub Releases. | Process Creation, File Creation | Execution | Medium |
| Detect processes making network connections to Steam Community profiles or Telegram channels followed by connections to unknown IPs, indicating dead drop resolution. | Network Connections | Command and Control | Low |
| Identify unexpected SOCKS5 proxy traffic originating from developer workstations. | Network Connections | Command and Control | Low |
Control Gaps
- Lack of strict application allowlisting for developer tools
- Inability to inspect encrypted traffic to legitimate services like GitHub, Steam, and Telegram
Key Behavioral Indicators
- Large 7z archives (78-167 MB) downloaded from newly created GitHub repositories
- Rust-compiled executables running from user directories
- Connections to Steam/Telegram for C2 resolution
False Positive Assessment
- Low
Recommendations
Immediate Mitigation
- Block known malicious GitHub repositories and URLs associated with this campaign.
- Isolate endpoints showing signs of Vidar stealer or GhostSocks proxy activity.
Infrastructure Hardening
- Enforce endpoint detection for Rust-compiled droppers and infostealer behaviors.
- Implement network monitoring for unauthorized SOCKS5 proxy traffic.
User Protection
- Instruct developers to install AI tools only from official, verified sources (e.g., claude.ai/install.sh or claude.ai/install.ps1).
- Deploy strict application control to prevent execution of unapproved binaries from user directories.
Security Awareness
- Educate developers on the risks of downloading 'leaked' or 'cracked' software from unofficial GitHub repositories.
- Establish clear organizational policies for the approval and installation of AI coding tools.
MITRE ATT&CK Mapping
- T1608.001 - Stage Capabilities: Upload Malware
- T1585.003 - Establish Accounts: Social Media Accounts
- T1566.002 - Phishing: Spearphishing Link
- T1204.002 - User Execution: Malicious File
- T1027 - Obfuscated Files or Information
- T1497.001 - Virtualization/Sandbox Evasion: System Checks
- T1555 - Credentials from Password Stores
- T1005 - Data from Local System
- T1102.001 - Web Service: Dead Drop Resolver
- T1090.003 - Proxy: Multi-hop Proxy
- T1041 - Exfiltration Over C2 Channel
Additional IOCs
- Urls:
github[.]com/realtime-voice-changer-app/realtime-voice-changer/releases/- Malicious GitHub Release URL distributing trojanized voice modifier archives.github[.]com/LTX-desktop/LTX-2.3/releases/- Malicious GitHub Release URL distributing trojanized video editor archives.github[.]com/nvidia-nemoclaw/NemoClaw/releases/- Malicious GitHub Release URL distributing trojanized AI tool archives.hxxps://github[.]com/leaked-claude-code/leaked-claude-code- Confirmed malicious GitHub repository distributing ClaudeCode_x64.7z.hxxps://github[.]com/my3jie/leaked-claude-code- Confirmed malicious GitHub repository distributing trojanized payloads.
- Other:
claude-cowork-win-x64.7z- Trojanized archive filename used in the campaign.opus-4-6-x64.7z- Trojanized archive filename used in the campaign.CopilotCowork_x64.7z- Trojanized archive filename used in the campaign.KawaiiGPT_x64.7z- Trojanized archive filename used in the campaign.WormGPT_x64.7z- Trojanized archive filename used in the campaign.NemoClaw_x64.7z- Trojanized archive filename used in the campaign.hyperliquid-bot_x64.7z- Trojanized archive filename targeting cryptocurrency users.bbg_free_x64.7z- Trojanized archive filename targeting finance users.idbzoomh- Known threat actor GitHub account.idbzoomh1- Known threat actor GitHub account.my3jie- Known threat actor GitHub account.