CVE-2024-3571: langchain-ai langchain-ai/langchain Path Traversal

Vulnerability summary

CVE-2024-3571 is a high (CVSS 8.8) Path Traversal vulnerability in langchain-ai langchain-ai/langchain. EPSS estimates a 1.87% chance of exploitation in the next 30 days, higher than 79% of scored vulnerabilities. 1 public exploit is recorded.

Timeline

1. 16-04-2024
   PublishedVulnerability first disclosed.
2. 01-08-2024
   Last modifiedVulnerability information updated.

Description

langchain-ai/langchain is vulnerable to path traversal due to improper limitation of a pathname to a restricted directory ('Path Traversal') in its LocalFileStore functionality. An attacker can leverage this vulnerability to read or write files anywhere on the filesystem, potentially leading to information disclosure or remote code execution. The issue lies in the handling of file paths in the mset and mget methods, where user-supplied input is not adequately sanitized, allowing directory traversal sequences to reach unintended directories.

• CWE-22Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal')The product uses external input to construct a pathname that is intended to identify a file or directory that is located underneath a restricted parent directory, but the product does not properly neutralize special elements within the pathname that can cause the pathname to resolve to a location that is outside of the restricted directory.