Intelligent System for Intrusion Detection Based on Machine Learning

Abstract

In today's digital age, security challenges threaten user privacy as networks face increased vulnerability to malicious attacks due to large data volumes. Intrusion Detection Systems (IDS) play a crucial role in identifying cyber-attacks and protecting system resources and users. This study utilizes machine learning classifiers (MLC) to analyze the NSL-KDD dataset, optimizing by preprocessing to remove irrelevant features. System performance is assessed using four attribute subsets, comparing model accuracy across DoS, Probe, U2L, and R2L attack classes to determine the best algorithm for each class. Using Random Forest with 20 features successfully achieved an accuracy of up to 99 % in intrusion detection.