Publication

Abstract : The constant development of new forms of malware has made it difficult to detect them using conventional approaches, but with the help of such sophisticated approaches as machine learning (ML). In this paper, the author aims at discussing the application of ML in improving malware detection and classification using supervised, unsupervised, and deep learning techniques. The survey focuses on important aspects of ML application such as feature selection and construction, and static and dynamic analysis of the application, as well as the interaction between ML and other cutting-edge technologies including edge computing and blockchain. The study shows how ML helps to mitigate zero-day threats and the advanced malware types namely polymorphic and metamorphic. It also provides information about drawbacks of ML which are adversarial attacks, data deficiency, problems related to scalability, and last one is regarding the model’s interpretation. In the last section of the paper, the author discusses the directions for future work, pointing out that the formation of datasets, evaluation procedures, and open-source platforms for reliable and effective anti-malware systems should involve joint work of researchers, businesses, and regulatory agencies.