基于谱聚类的社交网络差分隐私保护算法研究

A differential privacy protection algorithm in social network based on spectral clustering

针对权重社交网络差分隐私保护算法中噪声添加量过大以及隐私保护不均衡问题,提出了一种结合谱聚类算法与差分隐私保护模型的隐私保护算法SCDP。首先针对传统差分隐私保护算法直接向社交网络边权重添加噪声方式带来的噪声添加量过大的问题,结合谱聚类算法,将权重社交网络聚类成为不同的簇,对不同的簇采取随机添加噪声的方式,降低噪声的添加量,提高数据的可用性;其次设计新的隐私预算参数,根据社交网络边权重的大小决定噪声的添加量,实现更均衡的隐私保护;最后通过理论推导和实验证明了SCDP算法处理后的数据可用性更高。

Aiming at the problems of excessive noise addition and unbalanced privacy protection in the differential privacy protection algorithm of weighted social networks,a privacy protection algorithm combining spectral clustering algorithm and differential privacy protection model is proposed.Firstly,to solve the problem of excessive noise addition caused by the way of directly adding noise to the side weights of social networks by traditional differential privacy protection algorithms,combined with the spectral clustering algorithm,the weighted social networks are clustered into different clusters,and different clusters are randomly selected.The method of adding noise reduces the amount of noise added and improves the availability of data.Secondly,new privacy budget parameters are designed,and the amount of noise added is determined according to the weight of the social network side,so as to achieve a more balanced privacy protection.Finally,theoretical derivation and experiments prove that the data processed by the proposed algorithm have higher availability.