摘要：

针对软件定义网络(SDN)交换机和控制器的饱和攻击是SDN中的主要安全问题。在使用集成学习方法检测饱和攻击时, 现有方法通常使用距离或熵值等简单的信息计算方法修正证据, 可能存在信息丢失问题, 降低饱和攻击检测精度。为解决上述问题, 提出一种基于曲线决策融合的饱和攻击检测方法(SACOIN)。SACOIN首先计算多分类器概率矩阵的混乱程度修正多分类器内证据; 随后将多分类器概率矩阵转换为曲线并去除噪声, 提取重构小波的信号特征组成特征矩阵; 然后计算特征矩阵行内互信息, 基于上述互信息修正多分类器间证据; 最后使用D-S证据理论融合修正证据, 得到最终检测结果。实验结果表明, SACOIN在检测针对SDN交换机和控制器的饱和攻击时的准确率、精确率、召回率、F1值分别为92.3%、93%、92.1%、91.3%。

关键词: 饱和攻击, 软件定义网络, 滤波, 决策融合, D-S理论

Abstract:

Saturation attacks against Software Defined Network (SDN) switches and controllers are major security issues in SDN. When using ensemble learning methods to detect saturation attacks, existing methods typically use simple distance or entropy calculations to fix the evidence. This may lead to information loss, thereby decreasing detection accuracy. To overcome this problem, a novel curve decision fusion-based saturation attack detection method SACOIN is proposed. SACOIN first calculates the degree of confusion in the original probability matrix to fix the inter-multiclassifier evidence. The probability matrix of the binary classifier is then converted into a signal, and the noise is removed. SACOIN extracts the signal features from the reconstructed wavelet to form a decision matrix. Subsequently, the mutual information between the rows of the decision matrix is calculated to fix the original evidence of multiple classifiers. Finally, the Dempster-Shafer (D-S) evidence theory is used to combine the evidence and obtain the final detection result. The experimental results show that when detecting saturation attacks targeting SDN switches and controllers, SACOIN can achieve high accuracy, precision, recall, and F1 value of 92.3%, 93%, 92.1%, and 91.3%, respectively.

Key words: saturation attack, Software Defined Network(SDN), filtering, decision fusion, D-S theory