ANDROID APPLICATIONS MALWARE DETECTION: A Comparative Analysis of some Classification Algorithms

Abstract

The usage of the Android Operating System (OS) has surpassed all other operating systems and as a result, it has become the primary target of attackers. Many attacks can be geared towards Android phones mainly using application installation. These third-party applications first seek permission from the user before installation. Some of the permissions can be elusive evading the users’ attention. With the type of harm that can be done which include illegal extraction and transfer of the users’ data, spying on the users and so on there is a need to have a heuristic approach in the detection of malware. In this research work, some classification algorithms were tested to determine the best performing algorithm when it comes to the detection of android malware detection. An android application dataset was obtained from figshare and used in the Waikato Environment for Knowledge Analysis (WEKA) for training and testing, it was measured under accuracy, false-positive rate, precision, recall, f-measure, Receiver Operating Curve (ROC) and Root Mean Square Error (RMSE). It was discovered that multi-layer perceptron performs best with an accuracy of 99.4%.