Network Intrusion Detection System using Hyper Parameter Tuned Weighted XGboost Classifier

Abstract

Use of Internet of Things is increasing tremendously. With this increase in technologies like smart homes, smart devices and other automation, there is increase in malicious attacks. It is important to handle network security. Network intrusion detection system is vital to protect IoT devices from various malicious attacks. It includes dimensionality reduction, feature selection and classification. The dimensionality reduction is performed using principal component analysis (PCA) and the feature selection is accomplished using Ant Colony Optimization(ACO). Then, the classification is performed with the Extreme Gradient Boost Algorithm (XG-Boost). Benchmarked intrusion detection NSL-KDD dataset is being used and implementation is done on Colaboratory using Python. The system showed 98.2% accuracy, 99.2% precision, 97.1% recall and 98.1% F1-score, which are better than the existing intrusion detection system. The system proves to be robust based on experimental results in terms of evaluation of various performance metrics.