Enhanced GAN-Assisted Intrusion Detection Model for Addressing Data Imbalance Issues

With the introduction of IIoT systems, industry breakthroughs and linked smart objects revolutionize the world. The enemies are finding a lot of room to launch a number of devastating cyber assaults thanks to all this connection and dependability. However, implementing IDS systems may be tough due to the absence of dataset for training as well as testing. This makes it difficult to detect likely cyberattacks with the required accuracy. The method of detection might not be able to learn perfectly and detect cyberattacks significantly higher precision if there aren't enough records related to every threat category in the dataset. In addition, the detection of breaches procedure is becoming more complicated due to unbalanced datasets. In this light, our research suggests a more effective GAN-based IDS model for creating simulated data samples that can fool and evade security systems before directing the real-life network assault. To be more specific, it creates fictitious datasets and trains IDS on these in conjunction with the actual data in order to boost the system's efficiency. The specifics of the Improved Generative Adversarial Network-based Intrusion Detection System (IGAN-IDS) model that is suggested in this study are described in this section. With an accuracy of 99.6 percent, the suggested IGAN-IDS model surpasses all of its competitors, including BSA-IDS, GAN, MLP-GAN, Synthetic Minority Over-sampling Technique (SMOTE), and ADASYN. This proves that it outperforms all other approaches when it comes to classification.