AI hacking agents hunt exposed LiteLLM, Dify and Ollama installs for API keys

Started by Finley_27, Today at 12:17 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: AI hacking agents hunt exposed LiteLLM, Dify and Ollama installs for API keys   Views(Read 67 times)
Active members in this topic:
Finley_27(1)

Finley_27

Anyone running self hosted AI tools should check their security now. The Hacker News reports that Amazon's decoy MadPot servers recorded attacker AI agents probing internet facing AI apps for stored provider keys

Targets included LiteLLM, One API, Dify, Langflow, RAGFlow, Open WebUI, LobeChat and Ollama. The agents probably ran on ARTEX, an open source AI pentesting tool whose GitHub repo went offline on 8 October, though forks and a paid version remain

The FBI has issued a FLASH alert on ARTEX, and Amazon has moderate confidence the operators are linked to China. Cloudflare saw 400 database dump requests in one hour from an attacker using a compromised API token with no IP restrictions

Known flaws being used include Langflow CVE-2025-3248, fixed in 1.3.0, and RAGFlow CVE-2026-28797, fixed in 0.25.0. The advice is to change default admin keys such as LiteLLM's master key, rotate any exposed secrets, and add expiry dates and IP limits to API tokens

Self hosted AI tools often sit on the internet with default settings, which makes them easy pickings. Stolen API keys can run up huge bills or expose data. A quick audit of anything you have exposed is well worth doing this week

Do you run any self hosted AI tools? And have you changed their default keys?

Here more than I should be