Regular expression matching is playing an important role in deep inspection. The rapid development of SDN and NFV makes the network more dynamic, bringing serious challenges to traditional deep inspection matching engines. However, state-of-the-art matching methods often require a significant amount of pre-processing time and hence are not suitable for this fast updating scenario. In this paper, a novel matching engine called BFA is proposed to achieve high-speed regular expression matching with fast pre-processing. Experiments demonstrate that BFA obtains 5 to 20 times more update abilities compared to existing regular expression matching methods, and scales well on multi-core platforms.