作者: Chun-Ying Huang , Yi-Ting Tsai , Chung-Han Hsu
DOI: 10.1007/978-3-642-35473-1_12
关键词: Permission 、 Android malware 、 Naive Bayes classifier 、 Data mining 、 Android (operating system) 、 Malware 、 AdaBoost 、 Computer science 、 Support vector machine 、 Decision tree
摘要: It is a straightforward idea to detect harmful mobile application based on the permissions it requests. This study attempts explore possibility of detecting malicious applications in Android operating system permissions. Compare against previous researches, we collect relative large number benign and (124,769 480, respectively) conduct experiments collected samples. In addition requested required permissions, also extract several easy-to-retrieve features from packages help detection applications. Four commonly used machine learning algorithms including AdaBoost, Naive Bayes, Decision Tree (C4.5), Support Vector Machine are evaluate performance. Experimental results show that permission-based detector can more than 81% However, due its precision, conclude mechanism be as quick filter identify still requires second pass make complete analysis reported application.