DDoS 攻擊的防御是當前網絡安全研究領域中的難點。文章提出的方案能夠全面、主動預防DDoS 攻擊。使用蜜罐技術能夠預防已知類型的DDoS 攻擊；使用基于RBF-NN 能夠實時檢測DDoS 攻擊，檢測率可以達到97.9%；使用分塊包標記算法進行追蹤能夠保持較低的重構誤報率和較低的計算復雜度；使用分組漏斗的負載均衡技術能夠有效的緩解攻擊效果，保證了服務質量。該方案所采用的技術，實現簡單，代價小，具有很好的實際應用前景。
關鍵詞：分布式拒絕服務；拒絕服務；神經網絡；IP 追蹤

Abstract:The defense of DDoS attacks has become the difficult area in network security.The
proposed scheme can prevent DDoS attacks roundly and actively.It implements detector based on RBF-NN to detect DDoS attacks ,and the precise can reach 97.9%;implements an IP traceback scheme with packet marking in blocks,which provides higher precision and more efficient computation;implements a novel approach named Packet funneling to mitigate DDoS attacks,it ensures the quality of service. The technology of the scheme is practical and easy to
deploy,low price,it has the very good application prospect.
Key Words:DDoS；DoS；Neural Network;IP traceback