Anomaly based Intrusion Detection System using Hybrid Machine Learning Approach in IoT Environment

Abstract

To come up with a data-centric ecosystem, this work delves into the ideas of the Internet of Things (IoT) as well as machine learning (ML). The IoT and ML have become crucial tools for smart city solutions that depend on massive data gathering, processing, and decision-making. For the purpose of detecting and preventing intrusions into computer systems and networks, an intrusion detection system (IDS) keeps tabs on and analyzes data. Accurate and precise intrusion detection is difficult because network data has a large capacity and contains duplicate and unnecessary information. For the purpose of detecting network attacks in IDS, this study proposed a hybrid deep learning (DL) strategy including a convolutional neural network (CNN) and Xgboost. Additionally, this approach processes three publicly available datasets— such as the NSL-KDD, UNSW-NB15 as well as CICDDoS2019. This work preprocesses the obtained data using normalisation to preserve redundancy. Next, give the output from pre-processing stage to the feature selection procedure, enabling it to use principal component analysis (PCA) to compile the best features. Comparing the proposed hybrid strategy with existing methodologies demonstrates its efficiency.