COMPARATIVE ANALYSIS OF MACHINE LEARNING MODELS ON NSL-KDD DATASET

Abstract The security of modern communication networks cannot be sufficiently ensured without intrusion detection systems (IDS). The main objectives of these systems have been pattern recognition, signature analysis, and the detection of rule violations. Recent developments in machine learning (ML) and deep learning (DL) methodologies have shown potential as viable alternatives in the domain of network intrusion detection (NID). These methods can differentiate between typical and anomalous patterns. This study evaluates network intrusion detection systems (NIDS) using multiple ML algorithms, such as KNN, decision trees (DT), XGBOOST, and random forests (RF), with the NSL-KDD benchmark dataset. We analyze the precision, recall, accuracy, and F1 score of various ML techniques. The findings indicate that machine learning methods substantially enhance detection rates while minimizing false alarms compared to traditional approaches. This research demonstrates not only the feasibility of achieving a high detection rate of attacks but also the capability to make accurate predictions. These results clearly suggest that machine learning holds significant potential for the development of highly efficient NIDS systems. Keywords NSL-KDD, Machine Learning, Random Forests(RF), Decision Trees(DT), Network Intrusion Detection (NID), K-Nearest Neighbour (KNN), and XGBoost(XGB).