A CFS–DNN-Based Intrusion Detection System

The Internet communications are being developed to a great extent. With this development, an enormous amount of data is being generated. However, this data is not always secured. Intruders are always trying to misuse this data and gain unauthorized access, and hence, network security is also being compromised. An Intrusion Detection System (IDS) provides an efficient way to handle this. In this paper, an efficient IDS has been proposed which uses the NSL-KDD dataset which is a high-dimensional dataset. The dataset contains a large number of records, labeled as attack or normal. Correlation-based Feature Selection (CFS) method is chosen to select relevant and important features from the dataset for reducing the overall runtime of the proposed model, and a Deep Neural Network (DNN) classifier is used to examine if a record is normal or an attack. We tested our model using model validation and also compared the results with other existing models.