// Security Advisory Triage

pypdf memory exhaustion in /ToUnicode streams, fixed in 6.18.1

GitHub advisories — pip · published 2026-10-02 20:06 UTC
ACTION NEEDED PATCH_NOW SECURITY
WHAT CHANGEDGitHub published CVE-2026-102995 on 1 October 2026, rating it high severity. pypdf versions before 6.18.1 can exhaust memory when processing large /ToUnicode streams in PDF files. Version 6.18.1 patches the vulnerability.
WHY IT MATTERSServices that process untrusted PDFs can be forced offline by a crafted file. This is a follow-up fix, meaning earlier patches were incomplete and the exposure persisted.
WHAT TO DOCheck your dependency tree for pypdf < 6.18.1. If present, update to 6.18.1 in this cycle. If you process user-uploaded PDFs, prioritize this over routine updates.
Model-reported confidence 88% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+ghsa_id: GHSA-fp3h-c4fm-7vvf
+summary: pypdf: Possible large memory usage for large /ToUnicode streams (Follow-up 2)
+severity: high
+published_at: 2026-10-01T15:06:04Z
+cve_id: CVE-2026-102995
+vulnerabilities[].package.name: pypdf
+vulnerabilities[].first_patched_version: 6.18.1
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.5
triage reasoningNew high-severity advisory for pypdf with patched version. PDF processing is common in service dependency trees.
signalsnew advisory published · high severity rating · pypdf is a common PDF processing library in dependency trees · fixed version specified (6.18.1) · memory exhaustion vulnerability affects availability
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)