UTILIZING MACHINE LEARNING-BASED INTRUSION DETECTION TECHNOLOGIES FOR NETWORK SECURITY

Aug 4, 2026

Upload by Hazel

Effective intrusion detection systems (IDS) are increasingly vital for protecting computer networks as cyber-attacks grow in complexity. This work proposes an ML-based intrusion information detection approach named Stochastic Cat Swarm Optimized Privacy-Preserving Logistic Regression (SCSO-PPLR). Experiments evaluate the method using the KDDCup99 dataset, after Z-score normalization and feature extraction via PCA. Performance is assessed using accuracy, precision, recall, and F1-score, showing that SCSO-PPLR is an effective and acceptable IDS strategy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Intrusion detection systems and detection approaches

Network security challenges and IoT/WSN context

Motivation for improving IDS performance

Vol. 06, No. 1 (2024) 311-320, doi: 10.24874/PES.SI.24.02.014 Proceedings on Engineering Sciences [www.pesjournal.net](www.pesjournal.net) UTILIZING MACHINE LEARNING-BASED INTRUSION DETECTION TECHNOLOGIES FOR NETWORK SECURITY Rahul Kumar Sharma1 Arvind Kumar Pandey Bhuvana Jayabalan Preeti Naval Received 09.11.2023. Received in revised form 26.12.2023. Accepted 05.01.2024. UDC – 004.85 Keywords: Machine Learning, Network Security, Network attack, cybersecurity, Intrusion Detection Systems (IDS), Stochastic Cat Swarm Optimized Privacy-Preserving Logistic Regression (SCSO-PPLR). A B S T R A C T Effective intrusion detection systems (IDS) are becoming essential for maintaining computer network security due to the growing complexity of cyber-attacks. Machine Learning (ML) can increase the effectiveness of intrusion detection technology, which is an essential resource to safeguard network security. A novel ML technique for intrusion information detection called Stochastic Cat Swarm Optimized PrivacyPreserving Logistic Regression (SCSO-PPLR) is proposed. We assess intrusion detection systems using KDDCup99 dataset. The dataset is preprocessed using Z-score normalization to normalize the features. Next, Features are extracted by Principal Component Analysis (PCA). By comparing the results of the SCSO-PPLR methodology with traditional methods and using assessment criteria including accuracy, precision, recall, and F1-score, the model's performance is extensively evaluated. The study reveals that SCSO-PPLR is an acceptable strategy for intrusion detection in network security and it is effective. These insights broaden IDS and groundwork for further research on reliable cybersecurity remedies. © 2024 Published by Faculty of Engineering 1. INTRODUCTION An intrusion detection system (IDS) links to networks for suspicious behavior by scanning and examining them. In contrast, well-known tracking techniques are signature and anomaly-based detection, both of which the area of study on privacy has long investigated. Depending on the scope of use, intrusion detection systems can fall into several categories. For example, both host and network- based IDS, with capabilities ranging from single PCs to extensive networks, are among the most used. The ability to find undetermined dangerous code is constrained in the host-based intrusion detection system (HIDS), which is dependent on the single structure and watches critical operating system files for unusual activity (Sarker et al. , (2020)) . The scope of NIDS is represented in Figure 1. Figure 1. Framework of NIDS With the widespread usage of the Internet, network security has become essential. The prevalence of data accessibility has led to the emergence of serious risks, such as viruses and network intrusions that can cause significant damage to businesses. As a result, businesses are investing money in research employing clever strategies to enhance security, such as intrusion detection technologies. It gets more important topreserve research on intrusion detection in computer networks. The deployment of the IP protocol in version 6 (IPv6) raises serious concerns about network security, specifically about intrusion detection, as the IPv6 protocol allows for connections to the Internet of Things (IoT) (Da Costa et al., (2019)) . There are benefits and drawbacks regarding establishing security protection when using the security-by-design strategy, which is necessary due to the interconnected nature of the equipment. In addition, additional IoT techniques including the distributed ledger, cloud computing, machine learning (ML) and connected devices are considered. They are faced with the innate characteristic of resource and energy limitations present in IoT. Industrial control systems (ICS), smart grids, smart automobiles and medical gadgets are examples of cyber-physical systems (CPS) (Bertoli et al., (2021)) . Owing to IoT's explosive growth and the advent of w

FAQs

What problem does the paper address?

The paper targets the need for more effective intrusion detection systems to maintain network security amid increasingly complex cyber-attacks.

What method does the paper propose for intrusion detection?

It proposes Stochastic Cat Swarm Optimized Privacy-Preserving Logistic Regression (SCSO-PPLR), a machine-learning technique for detecting intrusion information.

How is the model evaluated in the study?

Evaluation uses the KDDCup99 dataset with Z-score normalization and PCA feature extraction, and compares performance using accuracy, precision, recall, and F1-score.

Aug 4, 2026

Upload by Hazel