Abstract: Distributed Denial-of-Service(DDoS) attacks are a major threat to availability of computer networks.In this paper,a novel scheme for early detection of DDoS attacks is proposed,which is involved with probability distributions of normal behavior on computer networks and DDoS attacks detection(model).The scheme employed statistical analysis of data from network connections to generate the pro-(bability) distributions of normal network connections.Based on the probability distributions,DDoS attacks detection model is presented.The feasibility of the scheme is validated through the simulated test.The(experimental) results show the effectiveness of our scheme in detecting DDoS attacks.Also,this scheme(provides) some directed sense for other network security detection research.