Enhancing Intrusion Detection with LSTM Recurrent Neural Network Optimized by Emperor Penguin Algorithm

Intrusion detection systems (IDS) have been developed to identify and classify these attacks in order to prevent them from occurring. However, the accuracy and efficiency of these systems are still not satisfactory. In previous research, most of the methods used were based on ordinary neural networks, which had low accuracy. Therefore, this thesis, with the aim of presenting a new approach to intrusion detection and improving its accuracy and efficiency, uses long-term memory (LSTM) optimized with the Penguin optimization algorithm (EPO). In the proposed approach, first, the features were pre-processed by normalization, cleaning, and formatting in number format. In the next step, the linear discriminant analysis (LDA) method was used to reduce the dimensions of the processed features, and after that, the EPO algorithm was used to optimize the size of the hidden unit of the LSTM network. Finally, the optimized network was evaluated using the NSL-KDD dataset, which is a widely used benchmark dataset in the field of intrusion detection. The results obtained for the training and test datasets were 99.4 and 98.8%, respectively. These results show that the proposed approach can accurately identify and classify network intrusions and outperform many existing approaches. Keywords: Intrusion Detection Systems, Penguin Meta-Heuristic Algorithm, Long-Term Memory Neural Network, Linear Detection Analysis.