Malware Evasion Attack and Defense

An adversarial example is an input sample which is slightly modified to induce misclassification in an ML Dataset Number of Samples classifier. In this work, we investigate white-box and grey-box Training Set 57170 (28594 clean and 28576 malware) evasion attacks to an MLbased malware detector and conduct Validation Set 578 (280 clean and 298 malware) performance evaluations in a real-world setting. We compare Test Set 45028 (16154 clean and 28874 malware) the defense approaches in mitigating the attacks. We propose a framework for deploying grey-box and black-box attacks to malware detection systems.