Detecting the Intrusion Attacks in MANET Using HGSO Based Feature Selection with Deep Recurrent Neural Network

In recent years, a Mobile Ad Hoc Network (MANET) has been defined as a Wireless Communication Network (WCN) composed of many highly mobile nodes traveling in var-ious directions. It alerts the service operations center when it discovers network-related behavior using an Intrusion Detection System (IDS). Nodes can connect at random in MANET since there is no infrastructure. However, compared to wired systems, ad hoc networks are more fragile, which has an impact on the features of MANETs. However, safeguarding MANETs against rogue nodes inserted arbitrarily into the network is a diffi-cult task. The paper suggests using a feature selection-based classifier to intrusions in MANET in order to address this issue. The preprocessed data is optimized using the Henry Gas Search Optimization (HGSO) method to provide an effective and efficient classification and prediction system for unexpected cyber-attacks. Secure transmission of data to other nodes is provided to counter intrusion attempts. Principal Component Analysis (PCA) reduces the number of dimensions in the data, improving the accuracy of the out-comes. Forward propagation uses a classifier called Deep Recurrent Neural Network (DRNN), which combines the unidirectional and bidirectional RNN, to determine if a re-sult is genuine or malicious. The effectiveness of the conventional and proposed models was compared using classification accuracy. The suggested model's classification accuracy is 96%, compared to the existing algorithms' 93% to 95% accuracy for the KDD-Cup99 dataset.