A Static Behavior-Based Method To Detect Malware On Android

Smart phone, with its powerful capabilities, has been wildly used in all areas of our daily life.The enormous kinds of applications installed in the mobile phone like weibo and mobile phonee-mail have changed the way people live and communicate.However, with the rising market share of Android smart phone, the malware(malicioussoftware) writers have began to target the Android system. The malware writers usually modify thepopular applications and embed the malicious code into those normal applications, then release themalicious package in the third-party application stores or forums. At present the main form ofmalwares are the premium-rate SMS-sending trojans which steal the users'property by registeringthe paid services. Besides, some malwares could steal the private data from the mobile phones.The android applications exist in the form of android package kit. For protecting the mobilephones, we must stop the malicious APK package from installing in the phones. It needs to detectthe APK package at important places like Android Market and user terminals and provide warningbefore the installation of malicious APK package. This thesis mainly research to use the staticmethods to classify the APK packages aiming at detecting the Android malwares. The mainaccomplishments are as follows:1. Show the malware's threat to the smart mobile phone. Analyze the advantage anddisadvantage of two main testing methods: signature-based detection and behavior-based detectionmethods. Summarize the research achievements of the testing methods at home and abroad.2. Summarize the security methods in the main part of circulation from APK developers tousers.3. Design a static detection method for the Android packages. Extract the class and functionname from the DEX document as behavior features and then study and classify them using fivemachine-learning algorithms provided by the WEKA tool. The simulation result shows that thismethod could effectively classify the Android APKs (The precision of every classification model isabove 0.7).