Evaluating Deep Learning Models for Android Malware Detection and Classification
Ahmed M. Neil, Eman Shaaban, Mirvat Al-Qutt, Karim Emara · 2025
Android applications play a crucial role in facilitating social interactions, conducting business operations and financial transactions, and monitoring health measures. However, their susceptibility to cyber threats has increased significantly with the rapid proliferation of smartphone usage. Malicious software can exploit application permissions to compromise these essential functions. Although Android usage and cyber threats continue to rise, applying deep learning to detect new Android malware is still a developing research area. In this paper we proposed four distinct deep-learning models using hybrid features(dynamic and static) to categorize Android mal-ware. Among these models, CNN-LSTM demonstrated superior performance, achieving an impressive accuracy rate of 98%. In addition, it compares the most promising results with relevant studies. The experimental evaluation results provide valuable insights into the effectiveness of these approaches.