Fake ChatGPT and Claude apps on GitHub spread Yoba malware
Original title: Attackers Hide Malware Behind Signed ChatGPT and Claude Binaries
Check Point says the two-month campaign sideloads a dropper through signed binaries; the payload seen so far is an XMRig miner.
Check Point said on October 27 that a campaign it calls Yoba has used fake ChatGPT and Claude desktop apps on GitHub to deliver malware. Releases impersonating GPT-5.6 Sol, GPT-6 Astra and Claude Opus sat in repositories with GitHub Pages, and some ranked highly in Google, which Check Point attributes to SEO poisoning. Signed OpenAI and Anthropic setup binaries loaded a malicious DLL, YobaDropper, which then installed YobaLoader. The loader uses Telegram and DNS-over-HTTPS to fetch the next stage. The observed payload was XMRig, but Check Point said the loader can deliver other executables or PowerShell.
Unconfirmed: Russian origin is based on indicators, and Check Point says later waves could deliver payloads other than XMRig.
- An earlier variant posing as the Cursor coding agent appeared in early June.
- YobaDropper checks the host process, administrator rights and keyboard layout.
- It skips Commonwealth of Independent States machines other than Ukraine.
- Persistence is a Scheduled Task running as Local System and disguised as a Microsoft task.
- Check Point cited indicators suggesting the actor has Russian origins.
Impact: Sideloaded loader can run a cryptominer or later payloads
- 1.Avoid installing AI applications from unverified GitHub repositories.
- 2.Restrict execution of software from unknown sources.
- ATT&CK
- T1574.002 DLL Side-LoadingT1204.002 Malicious FileT1053.005 Scheduled TaskT1102 Web ServiceT1496 Resource Hijacking
- Organizations
- Check Point
From the original article
2 min read at infosecurity-magazine.com
Attackers have used fake ChatGPT and Claude projects to distribute Yoba malware
By Alessandro Mascellino · infosecurity-magazine.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.