Enhancing Android Malware Detection through Filter-Based Feature Selection and Machine Learning Classification

Abstract

The escalating prevalence of Android malware poses a significant threat to cybersecurity. This research explores the effectiveness of filter-based feature selection techniques, specifically Information Gain, Chi-Square Test, and Fisher's Score, in analyzing Android permission patterns for malware detection. A suite of machine learning classifiers, including Decision Trees, K-Nearest Neighbors, Random Forest, Support Vector Machine, and Logistic Regression, were employed to evaluate the performance of these techniques. Results demonstrate that filter-based feature selection utilizing Information Gain and Fisher's Score outperformed the Chi-Square Test in terms of feature reduction, achieving classification accuracies of 91.53% and 91.22% respectively on the high-dimensional CICInvesAndMal2019 dataset. This study highlights the potential of filter-based feature selection methods, particularly Information Gain and Fisher's Score, for enhancing the efficiency and accuracy of Android malware detection.