An Intrusion Detection System Based on Data Analytics and Convolutional Neural Network in NSS-KDD dataset
D Kalaivani · 2022
Due to the internet's quick growth, intrusion attacks have been growing exponentially, making them a very important worry in the modern era. Cyber-attacks can target any of the millions of users of the internet, as well as international companies and government agencies. The creation of sophisticated algorithms to identify these network breaches is therefore one of the most important tasks in the field of cyber-security research. In order to recognise malicious traffic inputs, intrusion detection systems (IDS) are trained using data from internet traffic logs. Utilizing these techniques, malicious traffic inputs are detected. The most often used database for internet traffic record data is that maintained by the Network Security Laboratory's Knowledge Discovery and Data Mining (NSL-KDD) team. It also acts as the benchmark for present-day internet traffic. This framework seeks to discriminate between normal and abnormal (Denial of Service (DoS), Probe, User to Root (U2R), and Remote to Local (R2L)) categories in the NSL-KDD database with high detection precision and low false alarm rates. Several classifiers, including Naive Bayes (NB), K-Nearest Neighbour (KNN), Support Vector Machine (SVM), linear discriminant analysis (LDA), and Convolution Neural Network, will be used to achieve this (CNN). The unique and cutting-edge supervised detection techniques will be used in this study as the fundamental approaches to address the issue of the need for more labelled data during the IDS training process. The results of the trials show that, in terms of classification performance, the CNN classifier outperforms both recently presented approaches and other methods that are currently in use.