Adversarial Detection of Android Malicious Applications Based on Feature Decentralization
Qiang Gao · 2024
Machine learning-based methods for detecting malicious Android applications are widely researched, but their security is a concern. Adversarial attacks can easily evade the detection of these methods. This paper designs a feature-dispersion-based Android malware detection method that can effectively defend against adversarial attacks. The method takes the function call graph as features and reduces the feature dimensions according to the function types. The feature matrix is then separated to form 349 features of 1*349 dimensions, while 349 lightweight convolutional neural networks are used for malicious application detection. Experimental results on public datasets show that the method in this paper is effective in detecting adversarial attacks, with a high detection rate of 94.83%.