The limitations of network resource and the dispersion of network management are the two major difficulties for traditional networks to address the Distributed Denial of Service (DDoS) attacks. However, current defense methods are static and hysteresis, which are unable to locate the attackers accurately. Therefore, a dynamic defense using the two pivotal features, centralized control and dynamic management, of Software Defined Networks (SDN) is proposed. An OpenFlow-based switch shuffling model is built which employs greedy algorithm to remap user-switch link dynamically. After several shuffling, attacker could be differentiated from legitimate users and provide the latter with low latency uninterrupted services. The proposed approach is implemented in Ryu, the open source SDN controller, and the prototype is tested in a real SDN. The results of performance test show that with this approach attackers in limited times of shuffling can be isolated and the effects of DDoS attacks on legal flows can be reduced. The outcomes of defense ability test demonstrate that the efficiency of the proposed dynamic approach has nothing to do with the size of attack flow, but is only related to the number of attackers in the ring topology structure which is composed of a single controller.