Abstract: To solve the problem of low detection rate and long detection time of the existing DoS attack detection algorithm, a DoS attack detection algorithm was proposed based on higher-order statistics. The network traffic data packets were segmented and quantified in the algorithm. Followed, the characteristics of the accumulation was extracted which was applied to the detection of DoS attacks. By analyzing the 1998 DARPA intrusion detection data set, the algorithm can effectively detect DoS attacks. Compared with the traditional anomaly detection method entropy based on network traffic, the detection accuracy is greatly improved. In the time window of 1 s, the detection rate increases by 8%.