Using AI to Detect Android Malware Families
Saed Alrabaee, Mousa Al-kfairy, Mohammad Bany Taha, Omar Alfandi, Fatma Taher, Ahmed Hashem El Fiky · 2024
In today's digital era, many smartphone users often overlook security measures when installing apps, leaving Android devices particularly vulnerable to mal ware threats. Addressing this critical issue, there is a significant interest in leveraging Machine Learning (ML) and Deep Learning (DL) as proactive approaches for detecting and classifying Android mal ware, thus aiming to safeguard mobile and loT sectors. This study evaluates the effectiveness of data-driven methods in identifying and cate-gorizing Android malware families, specifically focusing on two advanced models: the 2-D Convolutional Neural Network (CNN) and Random Forest, which are essential for pattern recognition and decision-making. By utilizing a comprehensive dataset of Android malware, our research contrasts these models' performances and unexpectedly finds that Random Forest outperforms CNN, challenging the latter's reputed superiority in complex classification scenarios. This surprising result highlights Random Forest's efficacy in cybersecurity and underscores the potential of ensemble learning methods in this domain, suggesting new directions for future research in malware detection strategies. Our findings contribute to the cybersecurity field by enhancing understanding of ML and DL applications in malware detection and underscore the necessity for continuous exploration into more intricate scenarios and advanced learning methodologies to stay ahead of evolving cyber threats, especially within the Android ecosystem. This research not only opens new avenues for developing sophisticated and tailored MLIDL models but also significantly contributes to bolstering the security of mobile and loT devices, marking a significant step forward in the ongoing battle against malware.