Abstract: The time synchronization data filter estimative region is designed and a data filter algorithm based on linear regress is presented by using the distribution feature of the time bias and the linear feature of the time synchronization of node and the theory and the method of sample space estimation in probability statistics. This algorithm uses the estimative region to filter abnormal data and solves the problem that sensor networks are so easily attacked that the time synchronization is affected due to the high error rates of sensor networks. The algorithm is simple in decision process. The algorithm is avoided on the abnormal data. The computation costs are greatly reduced. The simulation result shows that this algorithm eliminates effectively the effect of the abnormal data on the estimated values of linear regress and increases the time synchronization accuracy.

Key words: wireless sensor networks, time synchronization, linear regress, data filter