Mutiple classifier system based android malware detection
Wen Liu · 2013
Smartphone becomes more popular recently. Much important information is stored and processed in the Smartphones. It attracts the attention of hackers. Malware is one of the most common security issues in Smartphone, especially for the Android system due to its compatibility. In this paper, we focus on the Android malicious application problem. As malwares with different purposes have different properties, only using a single classifier to discriminate the benign and malicious application may not be good enough. A detection method using Multiple Classifier System is proposed. Each base classifier is responsible for one type of malware. Android applications are classified as malwares when any one of base classifier decides they are malwares. The feature selection has been applied for each base classifier separately to increase the performance. The experimental results show that our proposed method outperforms than existing method PUMA, which classifies all types of malwares by a single classifier, in term of accuracy, TPR and FPR.