PROJECT TITLE :

Yes, Machine Learning Can Be More Secure! A Case Study on Android Malware Detection - 2017

ABSTRACT:

To address the increasing variability and class of modern attacks, machine learning has been widely adopted as a statistically-sound tool for malware detection. However, its security against well-crafted attacks has not solely been recently questioned, however it has been shown that machine learning exhibits inherent vulnerabilities that can be exploited to evade detection at test time. In different words, machine learning itself will be the weakest link in an exceedingly security system. During this paper, we depend on a previously-proposed attack framework to categorize potential attack scenarios against learning-based mostly malware detection tools, by modeling attackers with different skills and capabilities. We then define and implement a group of corresponding evasion attacks to completely assess the safety of Drebin, an Android malware detector. The most contribution of this work is the proposal of a simple and scalable secure-learning paradigm that mitigates the impact of evasion attacks, while only slightly worsening the detection rate in the absence of attack. We finally argue that our secure-learning approach will additionally be readily applied to different malware detection tasks.