Entropy-based clock skew measurements for mobile devices

摘要

Clock skew measurement can be realized in one-way communication by continuously collecting timestamps from the same remote device, and then find the slope of best fit to the offset dataset. However, offsets collected in mobile network may form multi-segmented distribution due to the change of base station or network adapter. This type of distribution usually cause the existing methods fail to provide accurate estimations. This paper introduces a new method to estimate the clock skew of remote network devices. In this method, a three-stage process is adapted to efficiently scan for the skew by which the entropy of adjusted offsets be minimum. Different from the existing approaches, this entropy-based method is not affected by multi-segment offset distributions and is able to deliver precise skew estimation with a short period of time. To compare our approach with existing ones, experiments of both classical and multi-segmented distributions have been conducted. The entropy-based method is the only one whose estimations be bound in 1 parts per million range.