Developing Explainable Deep Learning Models for Android Malware Detection
Sabrina Subah Nisa, Nafiz Al Asad, Abu Sayed Md. Mostafizur Rahaman · 2024
The Android operating system is becoming increasingly popular among mobile users due to its smooth handling and versatile features. But in recent years, Android malware has also increased due to its popularity. Cybercriminals deploy unwanted or dangerous software that might damage the device or steal personal data. Identifying and preventing malware threats in real-time is crucial to ensure security. In order to address this issue, we have created a framework that can continuously adjust to new mal ware threats while maintaining a high level of detection accuracy. We used a deep learning technique that can efficiently learn from both static and dynamic analytic character-istics of Android apps in order to accomplish this. Additionally, in order to determine which features are most relevant in identifying Android malware, we performed a feature importance analysis. This analysis gives security analysts and developers operative insights. Our experimental findings demonstrate that, with an accuracy of 95.87 %, our deep learning system performs better than several other methods.