Android’s wide adoption as the dominant mobile operating system has made it an attractive target for hackers, especially since users regularly keep sensitive personal data on their smartphones. This raises exposure to a wide range of security threats, emphasizing the importance of robust detection measures. This paper proposes an experimental methodology for identifying Trojan horse malware on Android devices through dynamic analysis, focusing on network traffic features extracted from a dedicated Trojan Detection dataset. In addition, a feature importance and stability analysis was performed using the J48 decision tree to enhance model interpretability and validate the robustness of the selected network traffic features. The dataset’s performance was assessed using multiple machine learning classifiers, where the J48 decision tree classifier achieved the highest detection accuracy at 99.0%.
Abdallah et al. (Thu,) studied this question.