A New Feature Selection Method Based on Dragonfly Algorithm for Android Malware Detection Using Machine Learning Techniques

Abstract

Android is the most popular mobile OS; it has the highest market share worldwide on mobile devices. Due to its popularity and large availability among smartphone users from all around the world, it becomes the first target for cyber criminals who take advantage of its open-source nature to distribute malware through applications in order to steal sensitive data. To cope with this serious problem, many researchers have proposed different methods to detect malicious applications. Machine learning techniques are widely being used for malware detection. In this paper, the authors proposed a new method of feature selection based on the dragonfly algorithm, named BDA-FS, to improve the performance of Android malware detection. Different feature subsets selected by the application of this proposed method in combination with machine learning were used to build the classification model. Experimental results show that incorporating dragonfly algorithm into Android malware detection performed better classification accuracy with few features compared to machine learning without feature selection.