RDE‐GAI‐IDS

The increasing presence of the Internet of Things (IoT) and edge devices in 5G and 6G virtual networks has created the necessity for inventive solutions to address challenges such as decentralized operations and security issues. Edge computing is a new emerging technology in cloud computing. This technology preprocesses and computes data on nodes near the edge computing network. Moreover, it also reduces the time complexity along with faster preprocessing of data. Despite all this, it providing proper security for the critical data generated from IoT has also become a challenge. As per the current trend, attackers are also using new techniques to trace network traffic, which can exploit the IoT device along with critical data and also lose data. Apart from this, Denial of Service (DoS) attacks and Distributed Denial of Service (DDoS) attacks are also common in edge computing networks. These attacks prove to be very fatal for IoT networks. Due to this, detecting them remains a major challenge. We proposed an Intrusion Detection System (IDS), namely RDE-GAI-IDS, by combining the ensemble learning (ENL) approach with the generative artificial intelligence (GAI) approach to detect attacks. For this, the selection of optimal features is done using a novel random forest feature importance (RFFI) approach. We used random forest (RF), decision tree (DT), extra tree (ET) and K-nearest neighbor (KNN) algorithms to build the Ensemble model. In addition, meaningful algorithms in GAI i.e. Generative adversarial networks (GAN) are also used to detect attacks on network traffic. We incorporated GAN in the proposed attack detection model to improve the robustness and diversity of data samples generated during training. The GAN model can better learn and generalize previously unknown attack scenarios and detect emerging threats in real time. To evaluate the proposed model, we used a new attack-based CIC-IDS-2017 dataset that contains almost all new attack-based traffic. The proposed model attains superior results with an accuracy of 99.918%, and minimum false positive rate (FPR) of 0.001%, and a detection time of 0.2ms. This paper proposes a unique attack detection model that uses ML methods and generative adversarial networks in distributed edge computing networks, and the improved numerical results demonstrate that it has the potential to make significant contributions to the IoT network security community.