Paper Title: MTAD: An Adversarial Defence Scheme for Securing Metaverse Satellite Networks
Conference Name: The 34th Joint Conference on Communications and Information (JCCI 2024)
Abstract: LEO satellites offer low-latency Internet connectivity for Metaverse but introduce security vulnerabilities in the presence of malicious adversaries. This study explores a white-box adversarial defense scheme for an AI-powered NIDS within LEO satellite networks. The proposed Moving Target Adversarial Defense (MTAD) policy makes adversarial perturbation techniques significantly expensive for attackers by returning decoy confidence scores of the model information. Overall, MTAD achieves a high detection accuracy of 98.42% while defending against perturbation attacks in Satellite NIDS models.