Executive brief
lobe-chat is an open-source AI chatbot application with a web-based interface. An attacker can exploit an unprotected /api/proxy endpoint to make unauthorized requests to internal services, potentially exposing sensitive data on the server's internal network without needing authentication. This allows reconnaissance and data exfiltration attacks against services normally isolated from the internet.
Technical details
The vulnerability is a Server-Side Request Forgery (SSRF) in the /api/proxy endpoint that fails to properly validate or restrict the target URLs. An unauthenticated attacker can craft POST requests to /api/proxy with arbitrary internal IP addresses or hostnames, causing the server to make requests on their behalf and return the response. The vulnerability exists in versions <= 0.150.5 and was fixed in version 0.150.6. No authentication is required; the attack vector is network-based with low attack complexity. An attacker can access internal services (CWE-918), exfiltrate sensitive data, and potentially pivot to internal infrastructure.
Affected products
- lobe-chat lobe-chat <= 0.150.5
Timeline
- 2024-05-10: disclosed
- 2024-05-10: patched: Version 0.150.6 released with fix