Adversarial Machine Learning for Robust Intrusion Detection Systems

EasyChair Preprint 14101

Abstract

Intrusion detection systems (IDS) play a crucial role in safeguarding computer networks against malicious activities. However, traditional IDS can be vulnerable to adversarial attacks, where attackers manipulate network traffic to evade detection. To address this challenge, researchers have proposed the use of adversarial machine learning techniques to enhance the robustness of IDS.

This paper provides a comprehensive review of the current state of adversarial machine learning for robust IDS. We begin by discussing the various types of adversarial attacks that can be launched against IDS, including evasion and poisoning attacks. We then delve into the different approaches proposed to mitigate these attacks, such as adversarial training, ensemble methods, and proactive defense mechanisms.

Furthermore, we explore the limitations and potential risks associated with adversarial machine learning techniques. We discuss the trade-off between detection accuracy and robustness, as well as the potential for attackers to adapt and launch more sophisticated attacks. We also examine the ethical considerations surrounding the use of adversarial techniques in IDS, emphasizing the need for transparency and accountability.

Keyphrases: Computer Networks, Intrusion Detection Systems (IDS), machine learning

@booklet{EasyChair:14101,
  author    = {Favour Olaoye and Peter Broklyn and Selorm Adablanu},
  title     = {Adversarial Machine Learning for Robust Intrusion Detection Systems},
  howpublished = {EasyChair Preprint 14101},
  year      = {EasyChair, 2024}}