Abstract:
This paper explores the applicability of evasion adversarial attacks, commonly used in computer vision, to network intrusion detection systems (IDS). It highlights the ke...Show MoreMetadata
Abstract:
This paper explores the applicability of evasion adversarial attacks, commonly used in computer vision, to network intrusion detection systems (IDS). It highlights the key differences between these domains, focusing on the feasibility of adding malicious perturbations to network packets. The study finds that the specific formatting and values required by networking protocols impose significant limitations on an attacker’s ability to modify data effectively. These constraints suggest that adversarial attacks in the networking domain may be less feasible and impactful compared to those in computer vision. Additionally, the paper discusses potential defenses that network IDS can employ to detect and mitigate such attacks.
Date of Conference: 10-13 November 2024
Date Added to IEEE Xplore: 30 December 2024
ISBN Information: