Android Malware Detection Based on Feature Selection and Weight Measurement

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