Efficient Feature Generation with Modified Whale Optimization Algorithm to Classify the Intrusion Detection

Abstract

In recent years, cyberattacks and network intrusions have emerged as significant threats to applications that are connected to the Internet of Things (IoT). Existing methods for preventing and detecting intrusions are not capable of accurately identifying every sort of attack or irregularity in network data.This is because to a number of constraints. Researchers have also proposed a large number of strategies that are based on machine learning; nevertheless, the effectiveness of these strategies in terms of classification accuracy or multi-class categorization is restricted. Through the utilization of a number of algorithms for the purpose of processing and filtering data, this study presents a data-centric technique for the detection of irregularities and intrusions. Improving the quality of the training dataset is accomplished by the utilization of the FGen-MWO Algorithm, which stands for Feature Generation with Modified Whale Optimization. K-Means is an additional application of automated machine learning that is used to find the method with auto-tuned hyper-parameters that is suitable for the most accurate classification of data. Not only does this approach reduce the computational cost of run-time data assessment, but it also generates an optimal algorithm that does not require any human tweaks to the hyperparameters. Overperforming preceding algorithms by a significant margin, the algorithm that was developed as a result can handle a multi-class classification problem with an accuracy rate of 99.7%.