面向SDN流表多模态感知与DRL协同防御DDoS方法OA

The centralized control architecture of SDN enhances management efficiency while posing risks of DDoS attacks.Addressing the challenges of traditional detection methods in handling covert attack behaviors within large-scale dynamic traffic and the tendency to mistakenly block short-term high-concurrency normal traffic,this paper proposed a DDoS defense system based on multi-modal deep reinforcement learning.This system achieved a dynamic balance between detection accuracy and re-source efficiency by integrating spatio-temporal feature decoupling and intelligent decision optimization.It maximized the avoi-dance of denial of service for non-attack traffic when resources are abundant.Experimental results show that the attack detec-tion accuracy rate averages 99.61％,with a false positive rate not exceeding 0.5％.This system reduces false positives for le-gitimate traffic while maintaining high accuracy,thereby ensuring the quality of network services during the defense process.