Research On The Software Security Detection Technology For Android Platform

In recent years, with the rapid development of mobile internet technology, the function and compatibility of intelligent mobile devices are increased gradually. More and more people choose to use Android mobile devices due to its great convenience. People’s daily life is gradually changing by the use of intelligent mobile devices. Meanwhile, the number of Android malware is increasing gradually, which result in a serious threat to the security of users’ information.From the perspective of Android malware detection, technology of software security detection on Android platform is discussed to solve the security problems of Android application in this thesis. The main contents are discussed as follows:1. In order to lay a foundation for the further research work, the theoretical knowledge of Android system structure and Android security mechanism is introduced, and the research achievements of Android software security detection at home and abroad are summarized in this thesis.2. The method of Android malware detection based on machine learning is discussed, both the method of Android file feature extraction and the method of feature selection are presented. In addition, the algorithm of feature selection based on Information Gain is deeply discussed in this thesis, and the method of Android malware detection using machine learning is presented.3. A model of software security detection on Android platform is proposed in this thesis, this model includes client and server, client is running on Android devices, server is used to detect the Android application received from the client. Both permissions and system API calls are extracted as Android application features, and the method of feature selection based on Information Gain is applied to select the Android application features. With these file features, the machine learning algorithms given in Weka are applied to train the classifier for Android malware detection.4. The results of experiments show that while using permissions and API calls as Android file features, Random Forest algorithm performs better, and the model proposed in this thesis can effectively detect Android malware.Both the method of extracting permission and API calls as Android file features and machine learning algorithms are applied in this thesis, this can not only detect Android malware, but also play a theoretical guiding role in Android software detection field. Therefore, this research has an important theoretical significance and practical value.