基于模式识别的自适应时钟重置检测算法

An algorithm for adaptive detection of clock reset based on pattern recognition

为解决网络测量中因测量主机的时钟重置导致单向时延测量误差的问题,根据时钟重置发生的不定时性和发生次数未知的特征,提出了一种基于模式识别的自适应时钟重置检测算法(PRBA)。该算法基于模式识别中的最大最小距离聚类算法对单向时延测量值进行聚类,同时利用时间序列技术中的一种低通滤波器过滤出噪声区间,从而有效识别网络行为特征,自适应地检测时钟值瞬时调整方式的时钟重置。实验结果表明,与现有同类算法相比,该算法具有较高准确性和自适应性。

In order to solve the problem of one-way delay measurement error in network measuring caused by clock reset, the paper proposes a pattern recognition based algorithm （PRBA） for detecting clock resets according to the uncertainty characteristics of clock reset occurrence. By clustering one-way delay measurement data based on the Batchelor and Wilkins Clustering algorithm in pattern recognition, and using a sort of low-pass filter in time series technology to filter out the ranges of noise, this algorithm can identify network behavior and finally detect multiple step phase adjustment in an adaptive manner. Number of experiments show that this algorithm is more accurate and adaptive than the existing algorithms.