Bugs that survive the heat of continuous fuzzing

This blog post examines why long-running continuous fuzzing (via OSS-Fuzz) still misses serious bugs, illustrating with three case studies—GStreamer (29 vulnerabilities), Poppler (including a 1‑click RCE in Evince), and Exiv2 (multiple CVEs). It identifies causes such as low fuzzer coverage, unfuzzed dependencies, lack of encoder testing, and bugs requiring large or long-running inputs, and presents a five-step workflow (preparation, coverage, context, value, triaging) and advanced techniques to improve fuzzing effectiveness.