Adversarial Machine Learning for Robust Intrusion Detection Systems

Abstract

In this study, adversarial machine learning to enhance IDS’s capability to counterattack sophisticated cyberattacks employed in the investigation. This paper describes challenges in practice of adversarial techniques, performance measurement and ethical issues. In the research proposal, the authors describe the comprehensive and multi-level method of detecting artifacts, building complex models, and gathering data. Researchers stressed important conclusions regarding aggressiveness of privacy-preserving methods, the need for developing new performance metrics, and the tension between robust model and detection performance. The research assists in developing IDS that are both efficient and formally correct in various contexts of a network.