Detection of ICMPV6 DDOS Attacks Using Ensemble Stacking of Hybrid Model-1 (CNN-LSTM) and Model-2 (RNN-GRU)

The internet, is a vital conduit for daily information exchange that operates on IPv6 and IPv4 protocols, facilitating connections but vulnerable to DDoS attacks, especially in ICMPv6 which is susceptible and inherited from IPv6 design. Despite A/IML advancements, attacks persist, causing significant losses. This research introduces two Deep Neural Network models: one merging CNN with LSTM, the other integrating RNN with GRU. Evaluated using datasets including NSL-KDD, alongside datasets from Sain Malaysian and Mendeley, resulting in accuracies of 80%, 97.01%, 95.06%, 72.89%, and 64.94%, respectively. Further showcasing the effectiveness of the approach by Ensemble Stacking using two sets of ICMPv6 datasets generated in distinct environments resulting in exceptional accuracies of 99.89% and 99.97% respectively.