基于EDLATrust算法的社交网络信息泄露节点概率预测

Social network information leakage node probability prediction based on the EDLATrust algorithm

在社交网络信息传播过程中,信息转发在用户之间广泛使用,但是存在着隐私信息在信息发布者未授权的情况下遭到泄露的问题。预测发现隐私信息泄露节点,对杜绝该类安全隐患具有重要意义。该文针对隐私信息泄露节点预测问题,提出了一种基于估计器的分布式学习自动机的信任推断(EDLATrust)算法,该算法能够推断社交网络中非直连节点之间的信任值,并减少算法收敛次数。基于信息转发时通常采用的线性传播和群传播2种典型传播模型,设计了满足信任传播模型的3种特征,采用XGBoost算法进行节点链接关系预测。该算法实现了对社交网络信息泄露节点的概率预测,利用该预测概率可以有效辅助推断信息传播过程中的信息泄露节点,从而提高了社交网络信息传播的安全性。在3个社交网络数据集上的实验表明,使用该算法能够有效地预测信息转发链当中信息的泄露节点,保护了用户的隐私安全。

Message forwarding is widely used in social network information systems. However, private information can be leaked without authorization from the information publisher. Privacy information leakage nodes need to be identified to eliminate such security risks. An estimator based distributed learning automata for trust inference(EDLATrust) is developed in this study to infer the trust level between non-directly connected nodes by reducing the number of convergence steps. The EDLATrust algorithm is combined with the XGBoost algorithm to identify privacy leakage in social network by using linear and group information transmission propagation models with three information dissemination characteristics. The algorithm predicts potential links in the information transmission chain and assists predicting information leakage points to improve the information dissemination security in social networks. Tests show that the model can effectively predict information leakage points in the information transmission chain for three real social network data sets to protect user privacy.