Towards Developing Generative Adversarial Networks Based Robust Intrusion Detection Systems for Imbalanced Dataset Using Hadoop-PySpark

The rise of machine learning has inspired the development of machine learning-based intrusion detection systems (IDS), yet their performance diminishes particularly with imbalanced data and missing datasamples. Generative Adversarial Networks (GANs) offer a solution by handling imbalanced data and addressing new vulnerabilities and challenges in cybersecurity and privacy for anomaly detection. GANs generate synthetic data to augment limited samples, enhancing the training of machine learning models to detect novel anomalies and patterns. This paper proposes an efficient GAN-based Random Forest (RF) model to introduce novel defense techniques with distributed computing using PySpark for big data processing because of computations efficiently in cluster of machines. Experimental results demonstrate that integrating synthetic data with existing datasets enhances model stability and robustness in detecting cyber attacks, with an accuracy of detection rate 99.94% when employing GANs generated augmented training dataset.