Maanak Gupta

This paper proposes an explainability-guided adversarial attack that successfully evades transformer-based malware detectors by perturbing the most influential components of the control flow graph representation.