Sliding Time Window Under The Conditions Of The Hilbert Transform Fast Algorithm And Its Applications In Network Traffic Anomaly Detection

Network traffic anomaly refers to the status that traffic behaviors depart from the normal behaviors,which has characteristics of a sudden attack and the unknown threatened characteristics. It is possible in a short period of time to network and network equipment to great harm. To detect anomaly rapidly and accurately and to respond to anomaly correctly has become the industrial sector and academic circles at home and abroad widespread concern the forefront of scientific issues, and how to improve the real-time detection is one of the key issues of anomaly detection.Research shows that the time-frequency analysis of traffic signals is an effective means of anomaly detection, and time-frequency analysis of the characteristics of one of the effective means is make Hilbert transform to traffic signals to produce its instantaneous parameters (instantaneous amplitude, instantaneous frequency, etc.). Use these parameters to detect abnormal network traffic, reduce traffic anomaly detection error rate and missed the seizure rate.Hilbert transform is usually made within a certain time window, in which the two adjacent window usually have some overlap (the part of the same data), and some double counting is in these two windows. Address the problem, this paper presents Hilbert transform fast algorithm under the conditions of sliding window, and applied to traffic anomaly detection. Main tasks include:(1) Introducing the research and analysis on Hilbert transform, according to the characteristics of traffic signals,the paper chose generalized Hilbert transform to make time-frequency analysis to improve the anti-noise capabilities of anomaly detection. In order to depicts clearly the distribution with time and frequency of the energy of the signal, we can also choose Hilbert-Huang Transform to make time-frequency analysis.(2) The traditional Hilbert transform need for a large number of Calculation in frequency domain,and it is difficult to use the overlapping properties of time domain window。We researched the Hilbert transform based window in time domain and proposed fast algorithm of Hilbert transform based window in time domain,increasing real-time of traffic anomaly detection processed by Hilbert transform. (3) The paper introduced the research and analysis on Hilbert-Huang transform,and proposed fast algorithm of Hilbert-Huang transform,increasing real-time of traffic anomaly detection processed by Hilbert-Huang transform.(4) We will put the two fast algorithm for network traffic anomaly detection. The fast algorithm gave rapid detection of abnormal flow through the amplitude analysis of variance of the instantaneous flow signal in history window and detection window.