A Classification Method of Social Network Members Based on Content Security

Abstract

With extensive and deep applications of Social Networking Services (SNS), more and more security issues are unfortunately related to it. Research shows unsuitable classification of social network members may induce misinformation and privacy leak. Thus, we propose a novel classification method of social network members based on content security. The method adopts LDA (Latent Dirichlet Allocation) to identify the topics of social networking content, and then takes topic vector as label to annotate the talking member. Finally, all the members are periodically classified according to topic labels. Moreover, an algorithm is also introduced to update the labels, so that the labels may be consistent in the trust decay. Preliminary experiments show that the method achieves 70%-85% customers' satisfaction.