Enhancing the Security by Analyzing the Behaviour of Multiple Classification Algorithms with Dimensionality Reduction to Obtain Better Accuracy

Abstract

The Intrusion Discovery System (IDS) is a critical component in the highly interconnected world of IT architecture. IDS, which may be implemented by a tackle or software bias and can reveal vicious conditioning in a networked environment, is regarded as a crucial component of security architecture. The Network Intrusion Detection (NID) system might be configured with machine learning algorithms to detect network threats quicker and more carefully. An efficient method is to evaluate various assaults using Dimensionality Reduction Algorithms. These algorithms are important because they extemporize point selection from big datasets. Additionally, this improves literacy speed. The effectiveness of network intrusion discovery systems for defense responses depends critically on speed. The experiment in this research makes use of the widely known benchmark CICIDS2017 dataset for intrusion detection systems. To find the best method and dimensionality reduction approach for this dataset, the dataset is given to a variety of classification algorithms. To decrease the amount of features in the data while safeguarding the important information, dimensionality reduction techniques like principal component analysis (PCA) are used along with classification algorithms like logistic regression, KNN, SVM, kernel SVM, random forest, and decision tree. Through extensive testing and assessment, the performance of various algorithm-dimensionality reduction pairs and assess each pair's correctness is ranked. By shedding light on the effectiveness of various tactics, the results of this study assist practitioners in selecting the optimal ones for similar situations. The end goal of this study is to provide information on the effectiveness of several classification algorithms with dimensionality reduction for intrusion detection tasks, assisting practitioners in selecting the appropriate approaches for similar scenarios. The source code of our paper available in this following link https://github.com/haasi003/intrusion-detection.git

Topics & Concepts