Machine intelligence approaches for preventing adversarial malware attacks in intrusion detection systems: A systematic review

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Abstract In academia, the epistemology of Adversarial Malware Attacks ( AMAs ) in the context of an Intrusion Detection System ( IDS ) has not been fully grasped. Using Machine Intelligence ( MI ), many attempts have been made to reproduce modeling techniques and methods within IDSs that can properly track, trace, and prevent AMAs from reappearing. However, there seems to be a lack of comprehensive and cohesive literature reviews in this area on which existing scholars can rely to pursue future research and broaden the field of Information Security and Networks ( ISN ) research. Motivated by the absence of a universal IDSs framework, termed a ’one-stop shop,’ this systematic review hopes to serve the research community by aligning thinking and firmly consolidating the historical knowledge and progress made in MI approaches, thereby providing a better understanding of AMAs, including their prevention and the implementation of viable and efficient IDSs. No other systematic review of this kind has yet been produced. Starting with nearly 1,000 papers and applying rigorous analysis, this study covers 132 research papers in multiple bibliographic databases since January 2020, highlighting studies on IDSs deployed for AMA detection and classification using MI learning techniques such as Machine Learning ( ML ) and Deep Learning ( DL ); methods involving feature extraction; studies discussing static, dynamic, memory, and hybrid feature analysis; and finally, studies providing experimental results and accuracy metrics from these IDSs. Throughout this comprehensive review, the emphasis is placed on highlighting the differences, strengths, and shortcomings from each MI-IDS approach, as well as providing discussions and suggestions for further exploration in future research. Following the PRISMA protocol, this systematic literature review (SLR) provides a rigorous, transparent, and reproducible foundation for knowledge by synthesizing all available evidence on IDSs over recent years.