A novel approach for mobile malware classification and detection in Android systems

With the increasing number of malicious attacks, the way how to detect malicious Apps has drawn attention in mobile technology market. In this paper, we proposed a detection model to seek and track malware Apps actions in such devices. To characterize the behaviors of Apps, dynamic features of each App were constrained in 166-dimension and a novel machine learning classifier is employed to detect malware Apps, and alarm will be triggered if an Android-based App is detected as malicious. With such, we can avoid a detected malware spreading out in larger scale, affecting extensively our society. Detailed description of the detection model is provided, as well the core technologies of this novel machine learning classifier are presented. From experiments performed on a set of Android-based malware and benign Apps, we observe that the proposed classification algorithm achieves highest accuracy, true-positive rate, false-positive rate, precision, recall, f-measure in comparison to other methods as K-Nearest Neighbor (KNN), Naive Bayesian (NB), Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), Decision tree (DT), Linear Discriminant Analysis (LDA) and Back Propagation (BP). The proposed detection model is promising and can effectively be applied to Android malware detection, providing early detection and the prospect of warning users of threatens ahead.