Android Malware Detection using TripleGuard Neural Network and Hybrid Bird Mating with Battle Royal Optimization

Android malware detection is a process that identifies and mitigates malicious software targeting Android operating systems, enhancing device security and reducing unauthorized access. However, it has potential disadvantages like false positives, advanced malware evasion, and the need for regular updates. To overcome this problem, this paper proposes a DL model with meticulous data preprocessing, eliminating missing records and standardizing numerical features through Z-score normalization. Feature extraction is then carried out to capture essential patterns within the pre-processed data. A unique hybrid optimization model called Hybrid Bird Mating with Battle Royal Optimization (HBMBRO), blending the Bird Mating Optimizer (BMO) and Battle Royale Algorithm (BRO), selects the most relevant features for optimal model performance. This study introduces a robust methodology for Android Malware detection, the "TripleGuard Neural Network" (TripleGuard NN), which amalgamates three specialized neural network components: The Optimized Autoencoder, Gated Recurrent Unit (GRU), and Artificial Neural Network (ANN). The synergy between the three neural network components offers versatile and robust Android Malware detection, with the Optimized Autoencoder identifying anomalies, the GRU analyzing sequential data for temporal Android Malware patterns, and the ANN delivering general Android Malware detection capabilities. The models within the TripleGuard NN are rigorously trained using MATLAB, and achieved an accuracy of 99.1%. This methodology promises a comprehensive and adaptable approach to Android Malware detection.