Anomaly Based Intrusion Detection System: A Deep Learning Approach
Sourou Tossou, Miftahul Qorib, Thabet Kacem · 2023
In recent years, computer networks have seen a considerable proliferation in terms of performance and total traffic volume. At the same time, cyber attacks have been on the rise ever since, which led to the emergence of Intrusion Detection Systems (IDSs) to deal with them. Conversely, artificial intelligence has been a popular technique that can be applied to a variety of purposes including detection of cyber attacks. However, most related work that leveraged artificial intelligence classifiers to address this problem used outdated datasets. In this paper, we implemented an anomaly-based intrusion detection system using deep learning algorithms with the goal of achieving higher performance while using a newer dataset. That is why we used the NSL-KDD dataset, which constitutes an improvement over the widely used KDD Cup 99 dataset, as it addresses some of its imperfections such as duplicated records and obsolete attack types. Then, we developed three deep learning classifiers that are Recurrent Neural Networks (RNNs), Multi-Layer Perceptron (MLP), and a hybrid Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) model. Also, we compared the effectiveness of our proposed model with Machine Learning classifiers such as Support Machine Vector (SVM), K-Nearest Neighbor (KNN), and Gradient Boosting (GB). Finally, we validate our findings with a performance evaluation of our model, which showed encouraging results.