Robust Android Malware Detection Against Adversarial Attacks

Android malware detection is one of the major concerns in the security field of Android mobile phones. Several researchers have experimented with and presented various techniques for Android malware detection using machine learning classifiers. However, in the current era of mobile security, developing machine learning classifiers for Android malware detection is not the only concern. The objective of such detection systems should be robust and include defence mechanisms against attacks that disrupt and tamper with the decisions of the machine learning classifiers. In our research, we implemented evasion and data poisoning attacks on the machine learning classifiers and analysed their performance. We have also applied defence mechanisms to the machine learning classifiers and studied their performance. The study concludes that the importance of adversarial machine learning and building robust Android malware detection are paramount.