A new detection method for LDoS attacks based on data mining

Abstract

The serving capabilities of networks are reduced by low-rate denial of service (LDoS) attacks that periodically send high-intensity pulse data flows. This type of attack shows a harmful effect similar to that of traditional DoS attacks, but their attack modes differ greatly. The high concealment of LDoS attacks makes it extremely difficult for traditional DoS detection methods to detect LDoS attacks. Meanwhile, the state-of-art detection methods for LDoS attacks have low-efficiency and resource-intensive and time complexity issues. We propose a novel detection method with analysis of abnormal network traffic under LDoS attacks that combines data mining technology. The judgement benchmarks were also established. The results from the experimental simulation on the simulated environment, physical environment and public datasets prove that the developed method can effectively detect LDoS attacks with optimal detection cost and low complexity, and has a high accuracy, a low false-negative rate, and a low false-positive rate.