Rat Swarm Optimization with Improved Gated Recurrent Unit for Intrusion Detection System

The Intrusion Detection System (IDS) has gained significant attention due to enhanced network utilization.However, various types of IDS approaches have been established in conventional research which focus on recognizing intrusions from datasets with the assistance of classification problems.However, the conventional techniques are unable to recognize malicious attacks due to the class imbalance issue.To overcome this issue, the Rat Swarm Optimization with Improved Self-Attention based Gated Recurrent Unit (RSO-ISAGRU) is proposed in this research for IDS classification.The RSO selects a set of best features by updating their positions based on their chasing and attack behavior.The weights are assigned by a self-attention mechanism which enables the ISAGRU to adopt attack patterns and enhance classification accuracy.The dataset is preprocessed by hash encoding and min-max normalization which convert the categorical feature into an integer format and normalizes the features.The accuracy, precision, recall and f1-score are taken as parameters for estimate RSO-ISAGRU performance.The RSO-ISAGRU achieves accuracy of 99.86%, 98.64%, 99.72%, and 99.83% for NSL-KDD, UNSW-NB15, CICIDS-2017, and CICIDS-2018 datasets when compared to ImmuneNet and Deep Neural Network (DNN).