What changed
PDFuzzer uses language models to extract format grammars and API relationships before solver-guided mutation, reporting up to 48% greater coverage and 31 zero-days across three PDF readers.
Why it matters
Structured document fuzzing can combine semantic format knowledge with executable search instead of relying on unguided mutation alone.
What to do
Prioritize vendor patches, isolate untrusted PDFs, and adapt grammar-plus-API fuzzing to parsers whose formats expose rich cross-object constraints.
Caveat
The ACM CCS paper reports coordinated disclosure, but no public implementation is available and deployed reader versions may already differ.