Amazon Uncovers AI Agent Campaign Targeting LiteLLM, Dify, and Other Apps for API Keys
Summary
Attackers are using AI agents to probe internet-exposed AI applications like LiteLLM, One API, and Dify for vulnerabilities that could allow them to steal API keys. Amazon's threat intelligence team captured records of these agents using decoy servers, revealing that stolen keys are then used to power the AI models searching for more keys.
IFF Assessment
The article describes a new attack campaign where AI agents are actively exploiting vulnerabilities to steal API keys, posing a direct threat to the security of AI applications and services.
Defender Context
Defenders need to be aware of the growing trend of AI-powered attacks targeting AI infrastructure. This campaign highlights the importance of securing API keys and implementing robust access controls for AI applications and models. Monitoring for unusual API usage patterns and suspicious AI agent activity is crucial.