An Efficient Intrusion Detection System Using a Multiscale Deep Bi-directional GRU Network to Detect Blackhole Attacks in IoT-based WSNs
E. Geo Francis, S. Sheeja, E. F. Antony John, Jismy Joseph · Journal of Multiscale Modelling · 2024
The IoT’s remarkable growth rate has attracted cybercriminals’ attention more than ever. The rising number of cyberattacks on IoT devices and intermediary communication mediums support the assertion. There are various ways for attackers to tamper with the IoT data distribution service. Since it affects data secrecy, the Blackhole attack was the most serious. For intrusion detection, deep learning algorithms are essential to IoT cyber security. Systems for detecting intrusions have been created to identify various attacks outside the firewall’s capability. The IDS categorizes the normal and abnormal aspects of the system based on its properties. In an integrated wired and wireless context, we developed a deep learning model to identify approved and unauthorized APs and detect Blackhole attacks in networks where raw traffic data has been gathered. To select the best feature sets exactly, we employ Sand Cat Swarm Optimization (SCSO), which tries to detect malicious attempts with the maximum detection rate possible. The Multiscale Deep Bi-directional GRU (MDBGRU) network also functions as a potent classifier, handling deep feature classification and threat detection in the complex IoT network environment. The outcomes of the suggested method’s detection accuracy are 99.27%, precision 89.34%, recall 84.32%, F1-score 77.31%, and Intrusion Detection Rate (IDR) 80.33%.