MalViT: An Approach to Enhancing Malware Detection

With the increasing ubiquity of computer systems and web services, the cybersecurity landscape faces an evergrowing challenge proposed by sophisticated and elusive forms of malware. Recent reports have highlighted the emergence of a new type of file-less malware that infects users' systems while leaving no obvious trace on the hard drives. Traditional methods namely the signature-based and heuristic detection have proven increasingly inadequate as malware creators continually devise novel evasion strategies. Hence, there is an urgent requirement for pioneering malware detection methods. In response to this escalating threat, this paper proposes a cutting-edge approach to malware detection, capitalizing on the transformative capabilities of the Vision Transformer (ViT) model. The malware files are converted into image representations and ViT is employed for the classification task, distinguishing between malicious and benign files. This approach not only promises to enhance detection accuracy but also addresses the pressing need for adaptability to ever-evolving malware variants.