MOBDroid2: An Improved Feature Selection Method for Detecting Malicious Applications in a Mobile Cloud Computing Environment

Abstract

This paper presents an ensemble machine learning (ML) based system for the detection of malicious applications in the Mobile Cloud Computing (MCC) Environment. The proposed system named MOBDroid2 applies a static feature analysis approach using the permissions and intents demanded by Android apps. The experiments conducted showed that the proposed system was able to effectively detect malicious and benign apps, achieving a classification accuracy rate of 98.16%, a precision rate of 98.95%, a recall rate of 98.20%, and a false alarm rate of 1.85%. The results obtained in our experiment compared well with other results reported in extant literature.