Malware Detection Using RGB Images and CNN Model Subclassing
Malware authors often upload their software to third-party application repositories to allow hackers to take control of a device by stealing passwords or providing access to contacts. Therefore, the development of an efficient malware detection tool is urgently needed. Malware detection researchers usually begins by extracting characteristics from specific sections of malware files, and this technique failed in case of zero-day malware. Unfortunately, malware classifica-tion remains a challenge, even if current state-of-the-art classifiers generally achieve excellent results, especially in image processing. To support efficient and powerful malware classification, we propose a CNN model subclassing architec-ture using RGB images. Malicious and benign files samples are converts to RGB images then the proposed classifier is able to recognize either it is malicious or not. We build our model using high-level API, then we examined out many opti-mizers. Finally, we got a Malware Detection Model as efficient as fast using RGB images, CNN from scratch with subclassing and Nadam optimizer. The end result we were given is 96% precision on a small database.
Further details : https://link.springer.com/chapter/10.1007/978-3-031-21101-0_1
Cite this work:
Ouahab, I.B.A., Alluhaidan, Y., Elaachak, L., Bouhorma, M. (2023). Malware Detection Using RGB Images and CNN Model Subclassing. In: Abd El-Latif, A.A., Maleh, Y., Mazurczyk, W., ELAffendi, M., I. Alkanhal, M. (eds) Advances in Cybersecurity, Cybercrimes, and Smart Emerging Technologies. CCSET 2022. Engineering Cyber-Physical Systems and Critical Infrastructures, vol 4. Springer, Cham. https://doi.org/10.1007/978-3-031-21101-0_1