社交网络中异常用户的识别与研究

发布时间：2018-06-04 22:45

本文选题：社交网络 + 异常用户可信度　；参考：《长春理工大学》2017年硕士论文

【摘要】：近年来,识别异常用户是社交网络中安全领域的重点研究内容之一。为了识别社交网络中的异常用户,本文首先提出了基于用户基本特征的异常用户识别方法,该方法利用G-N社区发现算法将用户分成孤立点和社区两类,结合粗糙集的理论计算用户的特征权重,依据特征权重计算有意义的特征信任值,结合特征权重和特征信任值计算用户可信度,所得可信度可用以识别异常用户。实验结果表明该方法随着数据集增大识别性能越好且识别效率越高。其次提出了基于用户间信任关系的异常用户识别方法,该方法依据用户间的信任关系特征计算信任关系度,引入衰减因子减少信任传播中衰减因素,结合信任链传播方式计算用户信誉度,根据用户信誉度判断用户类别。实验结果表明,该方法在已标记的数据集较少与异常用户数据比例不均衡两种情况下识别性能均很好,较传统的分类方法相比稳定性强、精确率和效率高。
[Abstract]:In recent years, identifying abnormal users is one of the key research areas in the security field of social networks. In order to identify abnormal users in social networks, this paper first proposes an abnormal user recognition method based on the basic user characteristics. This method uses the G-N community discovery algorithm to divide users into two types of isolated points and communities, combined with rough sets. It calculates the user's characteristic weight, calculates the meaningful characteristic trust value according to the characteristic weight, and calculates the user's credibility with the feature weight and the characteristic trust value. The reliability can be used to identify the abnormal users. The experimental results show that the method is better and the recognition efficiency is higher with the increase of the data set. Secondly, the user is based on the user. An abnormal user recognition method for inter trust relations, which calculates trust relationship according to the trust relationship between users, introduces attenuation factors to reduce the attenuation factors in trust propagation, calculates user reputation by combining the trust chain propagation mode, and judges the user class according to the user reputation. The experimental results show that the method is marked in the number of users. According to the two cases that the number of data sets is less than that of the abnormal users, the recognition performance is very good. Compared with the traditional classification methods, the recognition performance is stable, accurate and efficient.
【学位授予单位】：长春理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.08

【参考文献】