Benchmark
MalwareBench
Test whether AI agents can analyze obfuscated malware and extract useful evidence.
Explore the original evidence
Explore the original evaluation, its methods, and reported findings.
Open evalCoverage
Twelve initial cases across Windows, Linux, macOS, and Android. Published October 15, 2026.
Methodology
Agents investigate obfuscated binaries and produce a behavioral analysis and a reusable artifact extractor. Held-out malware variants test whether the extractor generalizes beyond the initial sample.