Collusion attack has become a serious threat to the evaluation system of e-commerce trust. In this paper, the genetic algorithm is used to find the global optimization of the two parameters of the support vector machine. And the standard SVM is transformed into the posterior probability output to estimate the probability of the sample. A reputation collusion identification model based on GA-SVM posterior probability output is proposed. Through the real conspiracy attack data validation, the proposed model can effectively identify the collusion attack, the area under the ROC curve can reach 0.954.%共谋攻击已成为电子商务信任评价机制所面临的严重威胁。本文通过采用遗传算法对支持向量机中的两个参数进行全局寻优,并将标准SVM的决策值转化为后验概率输出,对样本属于共谋的概率进行估计,提出了基于GA-SVM后验概率输出的信誉共谋识别模型。经真实共谋攻击数据验证,所提出的模型可有效识别共谋攻击,ROC曲线下面积能达到0.954。
展开▼