Denial of Service (DoS) attacks have continued to evolve and impact availability of the internet infrastructure. Many researchers in the field of network security and system survivability have been developing mechanisms to detect DoS attacks. This paper presents a novel dynamic entropy methodology for the study of DoS detecting. When an abnormal factor arises to agitate the current system the entropy must show an abrupt change. We examine the system using network traffic traces containing notorious DoS attacks and the results show that the proposed method can be able to detect anomalies with higher accuracy and lower false alarm rate.