Yoba campaign uses fake AI apps and signed installers to deliver miner

Key points

• Check Point Research followed Yoba for roughly two months.
• The operation used counterfeit GitHub repositories and digitally signed installers.
• Payloads included sideloaded malware and a cryptocurrency miner.
• Lures borrowed ChatGPT and Claude branding through fake AI applications.
• No CVE was involved; defenses should focus on behavior and supply-chain trust.

Check Point Research has spent roughly two months following an operation it calls Yoba. The campaign relied on counterfeit repositories on GitHub and installers carrying valid digital signatures, according to the research. Those components were used to place unwanted code on victim systems and then run a cryptocurrency miner.

Researchers describe the lure as AI-themed: fake applications that borrow the names of popular assistants, including OpenAI's ChatGPT and Anthropic's Claude. The use of recognizable AI branding gives the installers a plausible cover and can lower user suspicion. Check Point's findings tie the activity to the Yoba campaign rather than to a single vulnerability.

The practical risk is not a software flaw with a CVE. Instead, the operation abuses trust in code-signing and in developer platforms. Signed installers can bypass some reputation checks, while public repositories can host malicious projects that look legitimate. Once malware is sideloaded, the miner consumes endpoint resources and may be a precursor to other unwanted activity.

K-12 and government users are affected in the same way as any other target: users who download AI-branded tools from untrusted sources can introduce the payload. Because the campaign references mainstream AI products, staff and students searching for free or unofficial AI utilities are a plausible exposure point. The two-month tracking window suggests the activity is persistent enough to warrant monitoring rather than a one-off block.

Watch for new repository names, installer signatures, and AI-themed download pages. Defenders should compare signed binaries against allowlists, monitor for mining behavior, and treat unexpected GitHub-sourced installers as untrusted. Check Point's research provides campaign context, but organizations still need local detection and user reporting to catch variants.

What to do now

1. Require approval for unsigned and newly signed installers, and verify signer reputation before execution.
2. Restrict GitHub downloads on managed endpoints and route software requests through an approved catalog.
3. Hunt for cryptocurrency-mining indicators and unexpected child processes launched from installer or AI-app paths.
4. Add detections for AI-themed fake apps referencing ChatGPT or Claude, especially installers from public repositories.
5. Enforce application allowlisting and code-signing validation on staff and student devices.
6. Brief users to avoid unofficial AI tools and report AI-branded download prompts.
7. Review proxy and DNS logs for repository and installer delivery domains associated with the Yoba campaign.

Original source

Security Affairs