Security (4/5): xref parsing hardening — CVE-2026-22691, CVE-2026-27628, CVE-2026-41168#4
Open
icanhasmath wants to merge 3 commits into
Open
Security (4/5): xref parsing hardening — CVE-2026-22691, CVE-2026-27628, CVE-2026-41168#4icanhasmath wants to merge 3 commits into
icanhasmath wants to merge 3 commits into
Conversation
When startxref is broken (non-strict mode), _rebuild_xref_table scanned the whole file with a regex (b"[\r\n \t][ \t]*(\d+)[ \t]+(\d+)[ \t]+obj") that backtracks catastrophically on input containing long whitespace runs, exhausting CPU (CWE-1333 / CWE-400). Replace it with a manual byte scanner (_find_pdf_objects) that finds "<id> <gen> obj" markers in linear time using 1-byte slices (Py2/Py3 safe). Drop the now-unused `import re`. Mirrors upstream pypdf 6.6.0 (PR py-pdf#3594). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A malformed PDF whose cross-reference tables link /Prev in a cycle (xref A -> /Prev -> B -> /Prev -> A) made the xref-reading loop in read() follow /Prev forever, re-parsing the same tables (CWE-835). Track visited startxref offsets in a set and break (with a warning) if an offset repeats. Mirrors upstream pypdf 6.7.2 (PR py-pdf#3655). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A crafted /ObjStm with an enormous /N, or an xref stream with an oversized /Index (or /Size) subsection count, could force excessive read iterations (CWE-834). - _get_object_from_stream: clamp /N to (len(stream) // 3 + 1), the most "objnum offset" pairs the data can hold (raises in strict mode, warns and limits otherwise). Defensive: 1.28.6 also self-limits via the EOF parse error, but this matches upstream and bounds the loop directly. - _read_pdf15_xref_stream: clamp each /Index subsection count so the total cannot exceed (len(stream) // min_entry_bytes + 1) -- this is the substantive guard for the xref-stream iteration DoS. Mirrors upstream pypdf 6.10.1 (PR py-pdf#3733). Coverage note: the /N clamp's no-regression behaviour is unit-tested; a full binary xref-stream trigger for the /Index path is not constructed in tests (verified by inspection). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part 4 of 5 of the PyPDF2 1.28.6 security backport, targeting
1.28.6.x. Scoped toPyPDF2/_reader.pycross-reference parsing./Prevchains/Nand xref/Index//SizeBackported from upstream pypdf 6.6.0 / 6.7.2 / 6.10.1; Py2.7-safe. New tests:
Tests/test_security_xref.py(5, incl. a real cyclic-/PrevPDF). Validated under Python 2.7.18 —test_basic_features.py+ new tests pass, no regressions.🤖 Generated with Claude Code