Reproduce “MalwarePT: A Binary-Level Foundation Model for Malware Analysis” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Saastha Vasan, Yuzhou Nie, Kaie Chen, Yigitcan Kaya, Hojjat Aghakhani, Roman Vasilenko, Wenbo Guo, Christopher Kruegel and 1 more
MalwarePT provides a pretrained byte-level representation for malware analysis tasks. It uses byte-pair encoding and masked-token training, then adapts the encoder for API-call prediction, function labeling, and chunk-based file detection. On masked API-call prediction, it reaches 86.55% weighted F1, versus 64.54% from scratch. At a 0.1% false-positive rate on filtered Windows PE samples, it beats neural baselines but trails EmberGBM.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.