Android Malware Detection Based on Naive Bayes
Jiaqi Pang, Jiali Bian · 2019
Android phones are one of the most popular mobile intelligent terminals in the world. The open source nature of Android platform brings convenience to the development of thirdparty Android applications, but also provides conditions for Android malwares. In addition to Google's official app store, there are many mixed third-party Android app stores. Android users may download and install malware due to the large number of Android apps and the management omission of app stores, resulting in privacy disclosure, malicious fee deductions and other adverse consequences. In this paper, we propose an Android malware static detection method base on Naive Bayes. We extract requested permissions, system API calls, and the proportion of Activity among the four Android major components through Android packages. We use these three types of information as features to characterize each application, and perform classification model training and malware detection through Naive Bayes classifier. Our approach does not run Android programs, which reduces the cost of the experiment and completes the detection of malware before users' installation. The sample features we selected are common features that can be extracted from each Android application, so this detection method can be extended to all Android applications. We verify the feasibility of our method and achieve good results through the experiments on 6120 Android malwares and 6032 benign applications.