Comparative Analysis of Android Malware Classification Feature Selection on SVM Using PSO

Android operating system provides numerous applications within the Google Play service that can facilitate human activities. However, the lack of infrastructure security of some applications might cause malware infections. This situation potentially threatens the Android system and user data on mobile devices. To carry out this issue, the use of machine learning for malware classification has been popular in recent studies. However, as the preprocessing step, feature selection method was very important which can affect performance on machine learning. Therefore, this study proposed a comparative analysis using Particle Swarm Optimization as feature selection on the Support Vector Machine model for classifying Android malware. This experiment used CICMalDroid2020 dataset to compare Particle Swarm Optimization and Gain Ratio feature as a state-of-the-art method. The result showed that Particle Swarm Optimization was underperformed compared to the Gain Ratio method, achieved a 77.80% accuracy, 78.30% precision, 77.80<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> recall, and 77.35<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> F1-score