An Algorithm for Detection of Traffic Attribute Exceptions Based on Cluster Algorithm in Industrial Internet of Things
Lidong Fu, Wenbo Zhang, Xiaobo Tan, Hongbo Zhu
Abstract
With the focus on network security and the goal of meeting the requirements of fast speed and high accuracy of abnormal traffic detection of industrial Internet of Things, this paper proposes a hierarchical abnormal traffic detection method for the industrial Internet of Things. This method includes two abnormal detection methods: the first (crude) one detects traffic frequency based on statistical analysis; the second (sophisticated) one detects traffic attributes based on a clustering algorithm. The hierarchical detection method first detects the abnormal frequency of the network traffic. Property exceptions are then detected for suspected traffic. The paper first calculates the difference in the value of traffic frequency and smoothes it. Then, the exponential weighted moving average model is used to make the data conform to the statistical law, and the deviation correction of the model is proposed to reduce the error caused by the initial value. Focusing on the fact that a fixed threshold is not suitable for a dynamic network environment, a two-layer threshold interval method is proposed to reduce the rate of false alarms. For traffic attribute detection, this paper designs a clustering optimization anomaly detection algorithm for complex attribute feature data. The algorithm classifies the weighted distance and safety coefficient of the data according to the priority of the traffic attribute features, selects the data with the high safety coefficient as the clustering center, clusters the multi-feature data around the center, and applies it to the attribute anomaly detection. The simulation results show that the traffic frequency detection algorithm based on statistical analysis proposed in this paper can quickly detect the traffic frequency anomalies in the network. Moreover, the clustering optimization anomaly detection algorithm based on complex attribute features proposed in this paper can effectively detect the malicious attributes contained in network traffic and achieve a high detection rate and a low false detection rate to ensure the safety and reliability of the industrial Internet of Things.