An Efficient Machine Learning Based Models for Anomaly Detection in Network Traffic

Currently, anomaly detection in the network is a major concern of the day’s network clients in this age of advancement in technology. Network security has become paramount due to the increased number and volume of sophisticated computer crimes. There are many threats to the networks, and the most common are network anomalies that can result in system failures and inhibit the proper functioning of the networks. Though their efficacy in identifying network anomalies has been established, algorithms using Machine Learning and Deep Learning are still relatively unknown. This work’s objective is to thoroughly examine many methods for detecting network anomalies in Machine Learning. Using Machine Learning techniques on the NSL-KDD dataset-used for outlier identification-this study aims to enhance network intrusion detection. Comprehensive data preprocessing was undertaken to prepare the dataset for robust analysis. The study focused on multi-class and binary classification tasks, using Voting Classifier ensembles combining multi-layer perceptron with random forest and logistic regression with decision trees. The Voting Classifier models provided very high accuracy, whereas the multi-class model was found to have an accuracy of $\mathbf{99.} \mathbf{93 \%}$, while the binary-class model had an accuracy of 99. $\mathbf{82 \%}$, this makes the system very accurate, relying on the ensemble techniques to enhance the reliability and precision of an intrusion detection system.