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