IoT Malware Detection Using Deep Learning
Maurya Ranjeet Jeebodh, Niyati Baliyan · 2024
As digital infrastructure expands, the proliferation of malware files is escalating rapidly. Cyber-attacks have become increasingly complex and sophisticated, posing significant security challenges. Malware files can be challenging to detect when modified or obfuscated to appear different from benign files. Consequently, effective malware detection is crucial. This paper proposes an approach to detect malware, utilizing IoT Malware dataset for performance evaluation. The dataset comprises images derived from the byte code of malware files. Our approach leverages deep learning to detect malware by exploiting visual similarities between malware and benign samples. By extracting distinguishing patterns from the visual representations of these samples, our method efficiently categorizes malicious software. The model achieved a detection accuracy of $\mathbf{9 8. 2 9 \%}$, precision of $\mathbf{9 8. 8 3 \%,}$ F1-score of $\mathbf{9 9 \%}$, and a recall value of $\mathbf{9 9. 1 7 \%}$. These results demonstrate significant improvements in the precision, F1score and recall value compared to the previous best values of $\mathbf{9 8. 6 4 \%, 9 7. 1 2 \%}$ and $\mathbf{9 5. 9 7 \%}$, respectively. Experiments show that this image-based approach achieves high accuracy in malware detection, highlighting the potential of visualizing binary data for cybersecurity applications.