Abstract

The exponential growth of the internet and the proliferation of internet connected devices lead to the exhaustion of IPv4 addresses. To address this challenge IPv6 was introduced. ICMPv6 plays an important role in IPv6 and is prone to security vulnerabilities because it can be exploited for Distributed Denial of Service (DDoS) Attacks. Attackers can flood the network with the ICMPv6 messages to cause network disruptions. Henceforth, in our work, we introduce an Ensemble-Attention Hybrid Deep Learning Approach for detecting ICMPv6 messages flooding on IPv6 networks. An ensemble feature selection technique is incorporated, that combines filter and wrapper methods to identify essential features. To augment model precision, a transformer-based self attention mechanism is employed to ascertain attention weights assigned to the selected features. By concatenating the ensemble feature selection with the self attention mechanism, a Convolutional Neural Network (CNN) model is deployed to surpass the performance of existing methodologies. The experimentation on a benchmark dataset is carried out and the evaluation is based on metrics including False Positive Rate (FPR), detection accuracy, F-measure, recall, and precision. The proposed approach was evaluated on a benchmark dataset, achieving impressive performance metrics with a False Positive Rate (FPR) of 0.16%, detection accuracy of 99.87%, an F-measure of 99.85%, recall of 99.84%, and precision of 99.86%, demonstrating its effectiveness and reliability. Additionally, the findings indicate that the proposed approach surpasses the performance of existing methodologies.

Keywords: ICMPv6 DDoS attacks, Machine learning, Deep Learning, Self-attention