The Support Vector Machine Model and Application Based on Particle Swarm Optimization Algorithm

The support vector machine (SVM) model has been widely used in network security fields since it was proposed, but the original SVM model has some defects. In response to the limitations of Support Vector Machine models, numerous experts and scholars have proposed a variety of optimization methods, which have achieved favorable applications. The performance of a conventional SVM largely depends on the selection of its parameters, and manually selecting these parameters is often a time-consuming and inaccurate process. Based on this, this paper optimizes the penalty parameter and the kernel parameter of the SVM using Particle Swarm Optimization (PSO), and applies it to the anomaly type detection in the KDDCUP dataset. Focusing on the intrusion detection problem of the KDDCUP99 dataset, this paper establishes a comparison between the conventional SVM model and the PSO-SVM model, analyzing the classification accuracy and convergence rate of the two models on the test set. The results show that the parameters obtained through Particle Swarm Optimization allow the retrained SVM to maintain sensitivity to abnormal traffic while reducing false positives for normal traffic. This indicates, demonstrating the effectiveness and practicality of the PSO-optimized SVM model. Furthermore, its application in network intrusion detection is expanded.

I. Introduction

With the development of information technology and the advent of the digital age, the internet has become an indispensable part of people's daily life, learning, and work. However, concurrently, issues of cybersecurity have become increasingly prominent, posing serious threats and challenges to individuals, organizations, and society. Network intrusion is one of the most severe and common issues. Network intrusion refers to unauthorized access, tampering, or destruction of computer systems or networks, which can lead to data breaches, system paralysis, economic losses, and other multiple harms. In recent years, the frequency and severity of network intrusion incidents cannot be ignored. According to the monitoring data of the National Internet Emergency Center, phishing attacks account for as much as 44.9%, and ransomware attacks account for 30.7% of network attack incidents in China. As reported by the China National Internet Emergency Center, the economic loss due to data breaches in China exceeds 100 billion RMB annually. With the increasingly severe cybersecurity issues faced by China, network intrusion detection plays an increasingly important role in preventing network attacks and ensuring cybersecurity.