Security context

What an agent needs to avoid regressing past fixes and find the next vuln in this repo.

unstructured-io/unstructured

              main @ f6eea75
            

Fixes

CVEs

CRITICAL

Peak severity

46.2%

Coverage

Highlights

Denial of Service: 6 prior fixes. Scrutinize any change in this area.

Dockerfile: most-fixed (1 issue). Treat as high-risk during review.

6 high-severity fixes in this history; regressions here are high-impact.

Recurring patterns

The bug types that recur here, drawn from past fixes, not open vulnerabilities.

Path Traversal: Insecure handling of attachment filenames in MSG file processing allows attackers to write or overwrite arbitrary files via path traversal, as tracked under CVE-2025-64712. Prevention requires rigorous normalization and prefix-validation of attachment filenames.

Remote Code Execution: The ingestion and processing of natural language models via NLTK introduced risks of RCE due to untrusted archive extraction during downloading. This is mitigated by shifting to explicit, verified S3 downloads or alternative libraries like spaCy.

XML External Entity (XXE): The XML partitioning parser used lxml's iterparse without disabling entity resolution, exposing the application to XXE-based local file inclusion and Server-Side Request Forgery (SSRF).