Skip to Main Content
Abnormal network traffic has a very great harm to the network, so we need to quickly detect abnormal traffic. However, the existing detection methods take a lot of computational overhead, which will make it hard to meet the real-time requirement. This paper presents a distributed network traffic anomaly detection algorithm based on sliding window, which uses decomposable principal component analysis to handle network traffic signals. Through sliding time window, traffic anomaly detection will be limited to the specified scope of time. This significantly reduces the amount of data analysis to improve the speed of anomaly detection. Using the dataset from real network to simulate, we validate that the proposed algorithm is effective and feasible.