String matching algorithms have played critical role in many applications, such as DNA sequence comparison, network intrusion detection systems, and so forth. In this paper, we present a parallel multiple pattern matching method based on general purpose graphic processing units to realize fast string searching. In proposed method, we adopt a bit-parallel pattern comparison concept to accelerates string search to achieve efficient parallel search of multiple patterns of different lengths. In addition, we use CUDA framework to enhance the performance of searching string by leveraging GPU computing power. From the experimental results, the proposed method can achieve higher search throughput than other string matching methods. The proposed method is useful for genome sequence comparison and packet payload filtering.