Adversarial Attacks and Defenses in Deep Learning-Based Computer Vision Systems
Weijie Shan · 2024
Cyber Range is extremely vulnerable to widespread denial of service (DDoS) attacks from three malicious parties because it is a crucial component of the system that promotes digital space safety measures validation, system devices, weapons evaluations, attack safety dissatisfaction drills, and network threat assessment. It is also difficult for deep learning-based network intrusion detection systems (NIDSs) to defend against adversarial attacks, which are attempts by some adversaries to trick the classification/prediction algorithm by altering the input data. In this study, we suggested an adversarial training-based software-defined networking (SDN) detection and defence system. The system employs adversarial training to reduce the system's sensitivity to adversarial attacks and uses the Generative Adversarial Network (GAN) structure for detecting DDoS assaults. To conduct the study, we used the CICDDoS 2019 public dataset. The experiment results demonstrated that our technology could effectively detect common types of DDoS attacks when compared to alternative methods.