Executive brief
Crawl4AI is a web scraping library that can extract data from websites and PDFs. A vulnerability in its PDF scraping feature allows attackers to submit requests that download and process extremely large PDF files without any limits, consuming server resources (disk space, CPU, bandwidth) until the service becomes unavailable. In multi-tenant deployments, this can disrupt service for legitimate users.
Technical details
The vulnerability is an uncontrolled resource consumption flaw (CWE-400) in the PDFContentScrapingStrategy component. Root cause: the strategy is included in the list of untrusted-allowed types, and the non-streaming crawl endpoint permits untrusted clients to select it. When invoked, the PDF processor streams remote PDFs to disk with no size cap (relying only on a never-enforced content-length header), and iterates through all pages without limits. The default Docker configuration ships with no per-request wall-clock deadline. Attack vector is network-based with no authentication required; an attacker sends a POST request targeting a large or high-page-count remote PDF. Impact is denial of service only (no data disclosure or code execution). The vulnerability is fixed in version 0.9.3.
Affected products
- Crawl4AI Crawl4AI before 0.9.3
Timeline
- 2026-08-31: disclosed: GitHub Security Advisory GHSA-v2rm-hvrj-2x9q published
- 2026-09-15: advisory: CVE-2026-91941 published on NVD