LDoS Attack Detection Based on ASNNC-OFA Algorithm

Low-rate denial of service (LDoS) attack is a derivative denial of service (DoS) attack, which reduces the quality of service in the network in the means of sending high-strength and instantaneous streams of data in cycles. Based on the above attack, we proposed a method based on adaptive shared nearest neighbor clustering and outlier factor analysis (ASNNC-OFA) to detect it, whose core idea is to calculate the traffic characteristics of network. We divide these traffic characteristic data based on shared neighbor clustering, and use the outlier factor algorithm to perform anomaly analysis on the divided data by training threshold. The experiments were carried out on NS2 and test-bed respectively, the results of them show that the detection method we proposed has low false negative rate and false positive rate, so it can effectively detect LDoS attacks.