Unauthenticated Path Traversal in geyang/ml-logger

CVE-2025-10951 is an unauthenticated path traversal vulnerability affecting the geyang/ml-logger library in versions up to 0.10.36. The vulnerability exists within the ml_logger/server.py component, specifically within the log_handler and stream_handler functions. This flaw allows remote, unauthenticated attackers to traverse the file system and access unauthorized files on the hosting server. Publicly available proof-of-concept (PoC) code has been released, significantly lowering the barrier for exploitation. Defenders should prioritize updating instances of ml-logger to a secure version to mitigate the risk of unauthorized file disclosure or system compromise.

Impact

The vulnerability carries a CVSS score of 7.5, indicating a high severity risk. Successful exploitation allows an attacker to bypass file system restrictions to read arbitrary files, which may lead to the exposure of sensitive credentials, configuration files, or other proprietary information. Organizations utilizing ml-logger in web-facing server environments are at the highest risk, as the vulnerability requires no authentication or user interaction to exploit.

Recommendation

• Upgrade geyang/ml-logger to a version beyond 0.10.36 immediately to patch the vulnerable log_handler and stream_handler code.
• Review web server access logs for requests containing directory traversal sequences (e.g., ../, ..%2f) directed toward the ml-logger server endpoint.
• Restrict network access to the ml-logger server interface to trusted IP addresses if immediate patching is not possible.