pypdf CVE-2026-41313

The vulnerability requires loading a PDF file with a crafted large trailer /Size value while using incremental PDF update mode. Incremental mode is invoked when pypdf detects that a PDF has been appended to (a standard feature of the PDF format). The attacker must craft a PDF where the /Size value in the trailer dictionary is set to an arbitrarily large number, and the PDF must be supplied to pypdf in a context where it will be processed - this requires either direct user action to open the file in an application using pypdf or an automated system that processes PDFs without strict validation. No special configuration of pypdf is required; the vulnerability exists in the default parsing behavior. Exploitation is limited by the local attack vector (AV:L), meaning the attacker must have local file system access or the ability to convince a user to open the malicious PDF, and by the user interaction requirement (UI:P), which means the vulnerability cannot be exploited remotely or without user action.

This is a moderate-severity denial of service vulnerability with limited real-world impact. The CVSS 4.0 score of 4.8 reflects local attack vector (AV:L), low complexity (AC:L), and user interaction required (UI:P), with negligible confidentiality/integrity impact but low availability impact (VA:L). The local attack vector and requirement for user interaction to open a malicious PDF significantly reduces exploitability compared to network-accessible vulnerabilities. EPSS data is not provided, so exploitation probability cannot be benchmarked. No evidence of active exploitation (KEV status unknown; no CISA KEV listing in input). The attack requires deliberate user action to load a crafted PDF file, making this primarily a concern for applications that process untrusted PDF inputs automatically or where user education is weak. The low complexity and no-privilege requirements indicate that any user with access to the application can trigger the condition, but the local-only attack vector limits scope to on-premises or local-access scenarios.

An attacker crafts a malicious PDF file with a trailer /Size entry set to an extremely large value (e.g., 2,147,483,647). When a user or automated system opens this PDF using an application built on pypdf versions prior to 6.10.2, the library enters incremental mode and begins processing the artificially inflated object count, causing the application to consume excessive CPU and memory resources and become unresponsive for extended periods. Publicly available exploit code exists in the form of the patch itself, which reveals exactly how to construct such a malicious PDF.