An Enhanced Deep Neural Network Enabled with Cuckoo Search Algorithm for Intrusion Detection in Wide Area Networks

Abstract

A diversity of harmful software has been created as a result of the dramatic increase in internet usage, posing major risks to computer security. There is a great probability that the numerous computing operations performed through the network will be interfered with or altered, and as a result, effective intrusion detection systems are imperative. In addition, the attacks on the network are unpredictable, something that emphasizes the value of creating effective classification and prediction models. Machine learning (ML) and Deep Learning techniques have been used to evaluate datasets for intrusion detection systems (IDS). The employment of the DL-based approach enabled by feature selection helps to address challenges with data quality, handling high-dimensional data, and other related issues. Therefore, due to the large nature and volume of the IDS datasets, and the ability of DL-based models to learn categories incrementally through their hidden layer architecture to produce more accurate results in big data, this study proposes a Long-Short-Term-Memory (LSTM) model, and to further enhance the classification capacity of the projected DL method, the cuckoo search algorithm was introduced to select optimal features from the wireframe. The accuracy and subsequent detection of the suggested model positive and negative rates were evaluated. The experimental results show that the LSTM outperformed some other existing models with the highest classification accuracy of 99.7% and an error rate of 0.006.