SeqMobile: A Sequence Based Efficient Android Malware Detection System Using RNN on Mobile Devices
SeqMobile is an efficient system to detect Android malware using behavior-based sequence features instead of simple syntax features like permissions. It uses customized deep neural networks directly on mobile devices. By employing three performance optimization methods, the system achieves high detection accuracy (97.85%) while maintaining reasonable prediction times, proving that sequence learning can meet mobile device performance demands.