Android malware detection using the dendritic cell algorithm

Abstract

Most smartphones run on Android OS, which facilitates the installation of third-party applications. Unfortunately, malware also exists for the Android. Malware can perform various harmful activities. In this paper, we have collected the behaviors of 100 Android applications. These collected applications consist of 50 benign applications and 50 pieces of malware. The invoked system calls were logged to serve as the behaviors of these applications. Then, the data were input to the dendritic cell algorithm (DCA). The DCA was inspired by a danger model of the human immune system and is able to detect anomalies. We used the features of the DCA to perform anomaly detection and classified the collected applications as either benign or malicious. Our experiment results showed that the DCA could achieve a higher accuracy than either the decision tree, the naive Bayes, or the support vector machine.