Evaluating the Robustness of a Production Malware Detection System to Transferable Adversarial Attacks

As deep learning models become widely deployed as components within larger production systems, their individual shortcomings can create system-level vulnerabilities with real-world impact. This paper studies how adversarial attacks targeting an ML component can degrade or bypass an entire...