Android Malware Detection Based on Feature Selection and Weight Measurement

Abstract

With the rapid development of Android devices, Android is currently one of the most popular mobile operating systems. However, it is also believed to be an entry point of many attack vectors. The existing Android malware detection method does not fare well when dealing with complex and intelligent malware applications, especially those based on feature detection systems which have become increasingly elusive. Therefore, we propose a novel feature selection algorithm called frequency differential selection (FDS) and weight measurement for Android malware detection. The purpose is to solve the shortcomings of the existing feature selection algorithms in detection and to filter out other effective features. Weight measurement is used to optimize the detection accuracy of the classifier and improve the accuracy of detection. We combine the optimized features and the detection model for verification and evaluation. Experiments were conducted on the OmniDroid dataset, which is a large and comprehensive dataset of features extracted from 22,000 real malware and benign samples. Theoretical analysis and experimental results showed that the FDS algorithm and weight measurement are effective, feasible, and exhibit advantages over other existing malware detection models. In detecting Android malware samples, the proposed method can achieve an accuracy of 99% and an F1-score of 98%.