Efficient BERT-based Android Malware Detection

Abstract

Android smartphone malware poses a serious threat, which makes its detection more crucial than ever. Developing a reliable and effective method for Android malware detection remains a challenge in spite of academic and commercial efforts. The need for proactive malware defenses is growing as both individuals and organizations become more concerned about cyber dangers and malware attacks. A state-of-the-art class of attention-based deep learning techniques, Transformers, has shown impressive results. All things considered, BERT proved to be a viable way to fight Android malware. Its promise as a proactive protection mechanism against malicious software attacks is demonstrated by its capacity to outperform state-of-the-art alternatives. Furthermore, we test BERT on other datasets to gauge its effectiveness in various contexts. Our method has shown encouraging resilience against the quick growth of malware on Android systems in the ever-changing field of cybersecurity. With an accuracy of 99.09%, our model's performance is assessed on several datasets. Our findings show how successful our method is in proactively identifying malware risks.