Enhancing Real-Time Cyber Threat Detection Using Generative Adversarial Networks (GANs)

Abstract

The rapid evolution of “Internet of Things (IoT)” and 5G/6G technologies has amplified network interconnectivity, and also intensified exposure to sophisticated cyber threats. Advances in automated decision-making systems and remote data monitoring have resulted from the IoT and fifth and sixth generation (5G) technologies, which have enabled high communication rate. Some challenges arise including signature-based Network Intrusion Detection System (NIDS) struggles to adapt to zero-day exploits, risk in network traffic patterns, limited training, Inaccurate feature extraction and selection, Lack of cyber-attack detection. To tackle these issues, the research paper proposes an advanced real-time cyber threat detection framework integrating Generative Adversarial Networks (GANs) with Pigeon-Inspired Optimization and Tanh Function (SYN-GAN-PIA-TH) to enhance detection accuracy for both internal and external intrusions. The system architecture Processing (CEP) for automated preprocessing, and a (“Binary Gravitational Search Algorithm- Binary Grey Wolf Optimization”) with Linear Discriminant Analysis” (BGSA-BGWO-LDA) metaheuristic ensemble for optimized “feature extraction” and “dimensionality reduction”. “Long Short-Term Memory (LSTM)” models analyse temporal traffic patterns to mitigate bias and false alarms, while GSCV-SMOTE (Grid-Search Cross Validation with Synthetic Minority Oversampling Technique) addresses class imbalance in minority attack detection. Experimental results validate that synthetic data generated by GANs can effectively simulate diverse attack behaviors and achieving the high detection rate. The proposed approach establishes a robust, adaptive, and resource-efficient framework for real-time IoT network protection against emerging cyber threats. The results shows that proposed model outperforms significantly cyber threat detection using GAN, used metrics as detection rate, Number of Epoch, accuracy, Precision, F1-score and Recall.