Android Malware Detection Using Deep Transfer Learning Model for IoT Applications

The rapid increase in Internet of Things (IoT) devices has heightened concerns about Android malware threats, demanding more sophisticated intrusion detection methods. This paper introduces an innovative approach focused on pinpointing and classifying Android malware, specifically designed to address challenges within IoT applications. Leveraging Deep Transfer Learning, our model utilizes pre-trained neural networks to effectively extract intricate features crucial for distinguishing diverse malware types. This method ensures enhanced accuracy in identifying malware strains. By employing deep transfer learning, our model adeptly addresses the evolving landscape of Android malware threats, especially within IoT setups. Evaluation involved rigorous experiments on three established Android malware datasets—Malgenome, AAGM, and Drebin—emphasizing accuracy and F-score metrics. For the Malgenome dataset, it is observed that the proposed algorithm achieved an accuracy of 98.02% and an F-score of 96.99% at a dataset size of 20. For the AAGM dataset, the proposed algorithm achieved an accuracy of 95.09% and an F-score of 94.23%. For the Drebin dataset, the proposed algorithm achieved an accuracy of 97.89% and an F-score of 95.07%. Results unequivocally showcase our model's proficiency in accurately discerning and categorizing Android malware, even amidst varying dataset sizes. Significantly improved accuracy and F-scores were observed compared to traditional machine learning methods.