Enhancing Cybersecurity: A Hybrid Approach Using Generative Adversarial Networks and Advanced Oversampling for Attack Detection Model

This study was carried out to tackle the mounting challenge of data imbalance in intrusion detection systems which is a critical issue in the realm of information security. The issue is increasingly burdened by the complexity and rising frequency of cyber threats. This issue underscores the urgency for more sophisticated intrusion detection methodologies. We present a novel approach that synergistically combines Generative Adversarial Networks (GAN) with an Adaptive Synthetic Minority Over-sampling Technique (ASMOTE) to enhance detection and analysis efficacy in imbalanced datasets within intrusion detection frameworks. Utilizing extensive experiments with Random Forest and XGBoost models, we assessed the effectiveness of the GAN-ASMOTE methodology. The results demonstrated that GAN-ASMOTE markedly improved the recall rate and Fl-Score, especially in the precise prediction of minority classes, thereby affirming its enhanced capability to analyze imbalanced data with notable efficiency.