Enhancing Android Malware Detection: CFS Based Texture Feature Selection and Ensembled Classifier for Malware App Analysis

Smartphone usage is currently expanding at an explosive rate. Malware attacks can target these smartphones. Android smart devices contribute the major share in smartphones. So, malware authors focus on developing android malicious applications. In this study, a malware detection technique that is suitable for Android smart devices is proposed. It is a static technique based on visualization of android applications. In this proposed method, android applications are converted into grayscale images. Furthermore, the image’s texture features are retrieved using Grey Level Co-occurrence Matrix (GLCM) and Local binary pattern (LBP) techniques. Then feature filtration is performed using Correlation-based Feature Selection (CFS) to reduce the feature set. The number of extracted features is reduced to thirteen by performing feature filtration after feature extraction. It is observed that when the classification is applied to selected features, better results are obtained from filtered features rather than applying classification on features extracted through GLCM or LBP alone. This technique also reduced the prediction time of the detection system due to the small number of features.