A Particle Filter-Based Approach for Effectively Detecting Low-Rate Denial of Service Attacks

Abstract

Low-rate Denial of Service (LDoS) attack is one of great threats to computing platform and big data center. LDoS attacks are difficult to be found because its average attack rate is low enough to evade traditional DoS-based countermeasures. In this paper, the propagation back (BP) model of neural network is used to establish the nonlinear model of network traffic, and a method of LDoS attacks detection based on particle filter is proposed according to the mechanism of LDoS attack. In this method, the difference between the estimated value of the particle filter and the one step prediction is used as the detection basis, and a detection threshold is designed to determine the initiation and termination of the LDoS attacks for the purpose of detecting LDoS attacks. The method is verified in a test-bed network environment. Experiment results show that the detection rate reaches 98.2%, the false negative rate is 1.8%, and the false positive rate is 5.4%.