A Machine Learning Approach for Idle State Network Anomaly Detection

摘要

This paper proposes a Java application for detecting network anomalies due to DDoS attacks and congestion on a host in the idle state. It is also very challenging to detect and identify such problems especially when there is congestion in a network. The application uses parameters such as upload speed, download speed, number of packets transmitted and received, to analyse network traffic. The Multi-variate Gaussian technique has been used to detect anomalies in network traffic caused by DDoS attacks and congestion. However, in order to ensure that the anomalies detected over a specific interval of time are significant, t-tests have been used to test for their statistical significance.