ANOMALY DETECTION IN COMPUTER NETWORKS USING AI TECHNIQUES: А BENCHMARK STUDY

Analysis of network log files is crucial for identifying anomalies and addressing issues in information systems.The process of analyzing log files with traditional approaches can be highly demanding and tedious, especially when logs are large and unstructured.In research, machine learning algorithms and neural networks are applied to analyze network log files, aiming to detect specific system attacks.The UNSW-NB15 dataset, widely used in network security research, is employed to assess the prediction accuracy of various algorithms.Results indicate that algorithms like XGBoost, Random Forest, and Extra Trees achieve high accuracy rates exceeding 97%, while neural networks achieve approximately 96% accuracy.Training and prediction times for different algorithms are also analyzed, noting that neural networks, particularly LSTM, typically require longer training times.This study provides deeper insights into the performance of various anomaly detection algorithms in network log files and offers guidance for further research in cybersecurity.