Pattern matching is an important algorithm in intrusion detection system．The efficiency of this algorithm directly affects the intrusion detection system's performance. As in the process of traditional pattern matching algorithm, matches were done according to the character order in the pattern, a pattern matching algorithm based on statistics was proposed. By using the statistical characteristics of the string itself, the algorithm chooses the least frequency substring's characters in the main string as the key characters. According to this characters' position in the main string, jump matching are done. Thus the speed of pattern matching was accelerated greatly. The experiment results show that this algorithm is more efficient than other similar algorithms.