A New Data-Balancing Approach Based on Generative Adversarial Network for Network Intrusion Detection System

An intrusion detection system (IDS) plays a critical role in maintaining network security by continuously monitoring traffic and host systems to detect any potential breaches or suspicious activities. With the recent surge cyberattacks, there is growing need for automated intelligent IDSs. Many of these are designed learn normal patterns traffic, enabling them identify deviations from norm, which can be indicative anomalous malicious behavior. Machine learning methods have proven effective detecting payloads traffic. However, increasing volume data generated IDSs poses significant risks emphasizes stronger measures. The performance traditional machine heavily relies on dataset its balanced distribution. Unfortunately, many IDS datasets suffer imbalanced class distributions, hampers effectiveness techniques leads missed false alarms conventional To address this challenge, paper proposes novel model-based generative adversarial (GAN) called TDCGAN, aims improve rate minority while efficiency. TDCGAN model comprises generator three discriminators, with an election layer incorporated at end architecture. This allows selection optimal outcome discriminators’ outputs. UGR’16 employed evaluation benchmarking purposes. Various algorithms used comparison demonstrate efficacy proposed model. Experimental results reveal that offers solution addressing outperforms other traditionally oversampling techniques. By leveraging power GANs incorporating layer, demonstrates superior threats datasets.