MalAware: Effective and Efficient Run-Time Mobile Malware Detector

Abstract

Effective detection of malware is of paramount importance for securing the next generation of smart devices. Static detection, the preferred technique used so far, is not sufficiently powerful to defeat state-of-the-art malware, and will be even less effective in the near future. Dynamic malware detection guarantees better protection since it operates at run-time and can identify also unknown malware, however, the computational resources required are usually not affordable for battery operated devices. We propose MalAware, an effective, fast, and lightweight dynamic detection method. We identify malware by applying linear complexity classification algorithms to seven discriminating features and we improve the reliability of our detection using an efficient sliding windows mechanism. Our results, based on testing of about 2000 Android applications, demonstrate the timeliness and the effectiveness of detection in our approach. In fact, malware is detected within the first three minutes of execution with an F-measure of 0.85.