Detecting Malware on Android Devices Using CNN-LSTM with FDSWM based Feature Selection
S. Vasantha, Babburu Kiranmai, B. Jyoshna, Sri Ramakrishna, B. Suvarnamukhi, Sai Gokul Hariharan · Nanotechnology Perceptions · 2024
The widespread use of Android devices has made them attractive targets for cyberattacks. Traditional methods of detecting malware often struggle to keep up with evolving threats. In this research, a thorough analysis is undertaken to explore the potential contributions of deep learning and machine learning towards improving the detection of Android malware. Deep learning approaches based on GRU, LSTM, and CNN-LSTM networks are evaluated and compared with three machine learning algorithms: KNN, DT, and SVC. Using the CICAndMal2023 dataset, the effectiveness of these methods in distinguishing between malicious and benign Android apps is analyzed. The findings suggest that both deep learning and machine learning hold promise for improving Android malware detection. Specifically, the CNN-LSTM deep learning method demonstrated the highest accuracy of 99.7%. This indicates that deep learning techniques can provide a more reliable means of shielding Android users from evolving malware threats.