Long short-term memory-based intrusion detection system using hybrid grid search and sequential chimp optimization algorithm-based hyperparameter tuning

The current state of affairs in the cybersecurity situation depicts that there is a constant and rapid growth in the complexity and intensity of cyber threats, which calls for the need to design highly effective Intrusion Detection Systems (IDS). New trends in machine learning and optimization algorithms have been identified to improve IDS performance. However, the models of IDS that have been developed in the past do not perform well in the cases of imbalanced data, feature selection, and optimization. In this paper, we present an enhanced IDS utilizing the UNSW-NB15 dataset. To address specific challenges associated with Intrusion Detection Systems (IDS), we employ Recursive Feature Elimination (RFE) as a key strategy. The suggested model, called HGS-SChOA-Bi-LSTM, combines the Bi-Directional Long Short-Term Memory (Bi-LSTM) for threat categorization with the Hybrid Grid Search Algorithm and Sequential Chimp Optimization technique (HGS-SChOA) as the optimization technique. Various metrics are used to assess the model's execution, comprising the confusion matrix, F1 score, precision, accuracy, and recall. The results show how well our approach performs in comparison to current models. The suggested HGS-SChOA-Bi-LSTM model performs very well in intrusion detection, with 98.78% accuracy, 98.41% recall, 98.40% precision, and a 99.10% F1-score. The total training and prediction time is 89s, somewhat longer than for the other models; nevertheless, the increase in accuracy and performance rates justifies the length of time. The claim that the suggested model strikes a solid mix between recall and accuracy for cybersecurity applications is further supported by the high F1-score.