Malware Analysis Via Deep Language Models and Code Reuse

Discover the essential functions of the malware, and provide robust and accurate malware analysis.


  • Robust


    Fundamentally improves the robustness of the model without sacrificing accuracy. Robust against adversarial attacks, such as PE header modification, benign content appending, and binary randomization.

  • Self-explanatory


    Automatically select the most malicious functions for malware samples, and find all similar functions in our massive data for correlation analysis.