Enhancing IoT Network Security: Machine Learning-Based Intrusion Detection Using Vortex Search Optimization and SMOTE for Improved Classification
Chandrakanth Reddy Borra, Ramya Vani Rayala, Zabiha Khan, Srinivas Cheekati
Abstract
Internet security, and more particularly the prevention of harmful traffic on the increasingly popular Internet of Things (IoT) devices, has recently attracted a lot of attention. Such threats are hard to spot, which is why a sophisticated intrusion detection system (IDS) is required. As a smart IDS, machine learning (ML) shows promise in many domains, including the IoT. Features extraction models, however, should be used to extract data from the IoT environment so that ML models can use it. These models are crucial for the detection rate and accuracy. A number of network intrusion detection methods based on machine learning have been developed to successfully thwart these attacks on IoT networks. These methods typically employ feature extraction or feature selection techniques to reduce the dimensionality of input data before feeding it into machine learning models. This study employs a normalization pre-process on the IEEE Dataport input data and a synthetic minority oversampling technique (SMOTE) approach to select imbalanced data instances. Next, the Vortex Search Optimization Algorithm (VSA) is used to increase the classification accuracy by optimally selecting various features that have been extracted. The last step is for the input samples to be classified using a Fine-tuned based Support Vector Machine (FT-SVM). In order to determine how well the model works, it is compared to a standard data set. The experimental findings demonstrated the superior characteristics of the given model on the utilized dataset.