Smartphone applications (APPs) play essential roles in daily activities like online shopping, mobile banking, online transactions, etc. The exponential growth of online transactions has attracted the attention of hackers. Hackers are increasing their efforts to deploy malicious applications to users to steal sensitive information such as ATM PINs and bank account detail. The Traditional malware detection systems (MDS) require significant computational overload and time to analyze malware behavior patterns and find invasive tendencies. This research aims to expose the dangerous behavior of android malware to detect them quickly. We offer a negotiation by examining several types of static behavior patterns using BPSO (Binary Particle Swarm Optimization) to reduce computational complexity and pick the most optimal subset. The BPSO is hybridized with six machine learning techniques to get the complete solution for feature optimization and malware detection. The Binary Particle Swarm Optimization (BPSO) technique chooses an optimal subset from behavioral feature sets and provides the best fitness values. The Six machine learning techniques are utilized with BPSO to generate MDS models. The anticipated system has been empirically tested with three benchmark android datasets: DREBIN, MALGENOME, and the MENDELEY dataset. The proposed method achieved an accuracy of 96% with a 94% recall rate and 96 % f1 score. The high values of true positive (TP) and true negative (TN), indicate the model's effectiveness in both primary and secondary classes. The suggested technique has a meager computational cost, allowing for real-time application analysis.

Machine Learning, Android Feature Selection, Malware Detection, security, Binary Particle Swarm Optimization