面向蓄意攻击的网络异常检测方法

针对复杂网络受蓄意攻击频繁，而现有的检测方法大多忽略全局拓扑突变特征的问题.从网络全局拓扑的异常演化特征出发，提出网络路径相对变化系数(network path change coefficient，NPCC)r，量化节点间传输路径的变化.由斐波那契数列衍生出斐波那契演化域，用于区分正常和异常演化.将r作为核心度量参量，构建斐波那契演化域，形成网络异常检测方法，实现对异常的判定.结果表明，该检测方法的平均准确率为90%以上，高于最大公共子图(maximum common subgraph，MCS)及图编辑距离(graph edit distance，GED)的准确率，证明了所提检测方法的有效性.

Network Anomaly Detection Method for Intentional Attack

The anomaly triggered by intentional attack in complex networks is common but most existing detection methods ignore the global topology mutation feature. To solve this problem， based on the abnormal evolution characteristics of the global network topology， a network path change coefficient (r) was proposed to quantify the change of the transmission path between nodes. The Fibonacci evolution region was derived from the Fibonacci sequence to distinguish normal and abnormal evolution. r was used as the core measurement parameter to construct the Fibonacci evolution region， form a network anomaly detection method， and realize the determination of anomalies. The results showed that the average accuracy of the detection method is more than 90%， which is higher than the accuracies of MCS (maximum common subgraph) and GED (graph edit distance)， which proves the effectiveness of the proposed detection method.