摘要： 为确保对等网络节点交互的安全性，提出一种基于交易节点分类管理的网络安全模型。将失败的交易分为严重失败与一般不满意进行分类统计，以便更准确及时地检测恶意节点。在节点的直接交易过程中，根据交易历史记录，使用支持向量机分类器将网络中的节点划分为可信任节点、陌生节点和恶意节点，分别建立可信任节点列表与恶意节点列表，限制恶意节点的交易及反馈推荐行为。在反馈推荐意见统计表的基础上，利用Bayesian分类器对被评价节点进行分类，根据不同的可信度将可信任节点和陌生节点的反馈意见进行综合，再通过Bayesian估计调整节点的可信度。实验结果表明，与已有的安全模型相比，该模型对恶意行为具有更高的检测率，且交易成功率更高。

关键词: 安全模型, 对等网络, 节点分类管理, 支持向量机分类器, Bayesian分类器

Abstract: To ensure the security of the transaction in P2P network, a new network security model based on classification management of trading nodes is proposed. Through classification statistics of failure events in the trade between the local node and other nodes, the trade failure events are divided into malicious attacks, bad quality and so on, so that malicious nodes can be detected and controlled timely and correctly. According to transaction history records, Support Vector Machine(SVM) classifier is used to divide trading nodes into trust nodes, strange nodes and malicious nodes. The trust node list and the malicious node list are established to exclude the malicious nodes from trading. According to the statistical data of feedbacks from the other nodes, Bayesian classifier is used for the classification of the evaluated nodes. The model dynamically counts the feedback behavior condition, divided the feedback behavior into the honest feedback, the malicious feedback and so on. Experimental results show that compared with the existing trust model, the model proposed can obtain higher examination rate over malicious acts and the higher transaction success rate.

Key words: security model, P2P network, classification management of nodes, Support Vector Machine(SVM) classifier, Bayesian classifier

中图分类号: 

• TP393.08