dAEMD: Deep Autoencoder based Malware Detection from Android Network Flows

Android OS is an enticing target for attacks due to its popularity. Malware attacks are prevalent and growing. Further, the attack pattern is changing rapidly to avoid intrusion detection. Thus, effective malware detection that can adapt to rapid structure and behaviour changes is in demand. We provide a two-layer mobile malware detection method in this research. Deep learning represents the feature set into a latent feature space in the first layer, while the second layer is a straightforward multi-layer perceptron classifier. Elastic Weight Consolidation was added to the neural network classifier to enable continuous malware learning. We ran trials to evaluate performance. We compared our model’s accuracy to other machine learning models. We used cutting-edge methods to build our framework. Performance comparison with the state-of-the-art approaches shows the efficacy of the proposed framework. It can also learn new threats while maintaining detection performance. The testing findings show that our framework can detect intrusion with 98.8% Precision and 98.7% Recall. Additionally, the continuous learning system can accurately learn malware.