基于支持向量机和多资源最大最小公平的DDoS防御

Multi-resource max-min fairness and support vector machine based DDoS defense

采用分布式过滤的方法防御分布式拒绝服务(DDoS)攻击,通过将分布式防御合作限定在互联网自治域(AS)内,为应对选取了合适的网络范围,且考虑了带宽和受害机处理能力这2类资源及其相互作用.基于支持向量机(SVM)的多资源最大最小公平(SMMF)算法,根据受害端流量情况动态调整自治域边界的过滤器参数,保证了多资源最大最小公平,以达到较优的防御效果.模拟实验表明,该算法在具一般性的攻击场景下能有效抑制攻击流量,且在已有方法失效的情况下仍能保证合法流量吞吐量维持在正常水平.在路由器上实现了该过滤器,结果表明,即使安装上千个过滤器也只需极少量的内存,且仍能保证路由器的正常吞吐率.

Distributed denial of service （DDoS） attack was defended by distributed filtering.Distributed defense was restricted inside autonomous system （AS）,which was a suitable bound for defense.Both bandwidth and processing capability of victim were considered.The filtering threshold was dynamically adjusted in AS edge according to the throughput of victim in support vector machine （SVM）-based multi-resource max-min fairness （SMMF） algorithm.Then SMMF achieved multi-resource max-min fairness and was much effective.Simulation results demonstrate that attacking traffic can be depressed in a common scenario and the legitimate throughput can be kept in a normal level when current methods fail.A realization of filters on PC-based router indicates that only a very small amount of memory is needed and the packet throughput is still normal when thousands of filters are installed.