Spectral Graph Convolutional Neural Network for Decentralized Dual Differential Privacy

doi:10.3969/j.issn.1671-1122.2022.02.005

Abstract

Abstract:

Graph convolution neural network is a multi-task oriented and widely-used deep learning model. This paper focused on the protection of node relationship information and node feature information of graph convolutional neural network in spectral domain for decentralized scenes, and proposed a spectral graph convolutional neural network based on dual differential privacy protection mechanism called DDPSGCN. Given the total amount of privacy budget, the Laplacian mechanism and Gaussian mechanism are allocated privacy budget, and the parameters of the two distributions are estimated by privacy loss and Chernoff bound theory. The paper proposed a graph convolution neural network training algorithm based on block chain decentralized differential privacy processing mechanism under the influence of two kinds of distributed noise. Experiments show that the decentralized dual differential privacy mechanism can ensure the privacy of the original data without leakage under the premise that the accuracy of semi-supervised node classification task is reduced by less than 1%,which has higher privacy protection efficiency and stronger robustness against attacks compared with the single privacy protection mechanism.

Key words: dual differential privacy, decentralized differential privacy, spectral graph convolutional neural network, blockchain

CLC Number:

TP309

LIU Feng, YANG Chengyi, YU Xincheng, QI Jiayin. Spectral Graph Convolutional Neural Network for Decentralized Dual Differential Privacy[J]. Netinfo Security, 2022, 22(2): 39-46.

Figures/Tables 3

References 17

[1]	HE K M, ZHANG X Y, REN S Q, et al. Deep Residual Learning for Image Recognition[C]//IEEE. 2016 IEEE Conference on Computer Vision and Pattern Recognition, June 27-30, 2016, Las Vegas, NV, USA. New York: IEEE, 2016: 770-778.
[2]	GEOFFREY H, LI Deng, DONG Yu, et al. Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups[J]. IEEE Signal Processing Magazine, 2012, 29(6):82-97.
[3]	YING R, HE R, CHEN K, et al. Graph Convolutional Neural Networks for Web-scale Recommender Systems[C]//ACM. 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, August 19-23, 2018, London, UK. New York: ACM, 2018: 974-983.
[4]	KIPF T N, WELLING M. Semi-supervised Classification with Graph Convolutional Networks[C]//IEEE. 5th International Conference on Learning Representations, April 24-26, 2017, Toulon, France. New York: IEEE, 2017: 1-14.
[5]	XU Bingbing, CEN Keting, HUANG Junjie, et al. Survey of Graph Convolution Neural Networks[J]. Chinese Journal of Computers, 2020, 43(5):755-780.
	徐冰冰, 岑科廷, 黄俊杰, 等. 图卷积神经网络综述[J]. 计算机学报, 2020, 43(5):755-780.
[6]	KASIVISWANATHAN S P, LEE H K, NISSIM K, et al. What Can We Learn Privately?[J]. SIAM Journal on Computing, 2011, 40(3):793-826. doi: 10.1137/090756090 URL
[7]	MIRONOV I. Rényi Differential Privacy[C]//IEEE. 30th IEEE Computer Security Foundations Symposium, August 21-25, 2017, Santa Barbara, CA, USA. New York: IEEE, 2017: 263-275.
[8]	LI Q B, WU Z M, WEN Z Y, et al. Privacy-preserving Gradient Boosting Decision Trees[C]//AAAI. 34th AAAI Conference on Artificial Intelligence, February 7-12, 2020, New York, NY, USA. Palo Alto: AAAI, 2020: 784-791.
[9]	REUBEN J. Towards a Differential Privacy Theory for Edge-labeled Directed Graphs[EB/OL]. https://gi.de/ueber-uns/historie, 2020-07-19.
[10]	XIAO Biao, YAN Hongqiang, LUO Haining, et al. Research on Improvement of Bayesian Network Privacy Protection Algorithm Based on Differential Privacy[J]. Netinfo Security, 2020, 20(11):75-86.
	肖彪, 闫宏强, 罗海宁, 等. 基于差分隐私的贝叶斯网络隐私保护算法的改进研究[J]. 信息网络安全, 2020, 20(11):75-86.
[11]	PINOT R, MORVAN A, YGER F, et al. Graph-based Clustering under Differential Privacy[EB/OL]. https://xueshu.baidu.com/usercenter/paper/show?paperid=52f1b38f2a9d4bab19fd71507a63b14e&site=xueshu_se, 2020-06-15.
[12]	YIN Yiping, LIAO Qing, LIU Yang, et al. Structural-based Graph Publishing Under Differential Privacy[C]//IEEE. 8th IEEE International Conference on Communications, August 11-13, 2019, Changchun, China. New York: IEEE, 2019: 67-78.
[13]	GAO Tianchong, LI Feng. Protecting Social Network with Differential Privacy Under Novel Graph Model[J]. IEEE Access, 2020, 20(8):276-289.
[14]	SEALFON A, ULLMAN J. Efficiently Estimating Erdos-Rényi Graphs with Node Differential Privacy[C]//MIT. 32nd Advances in Neural Information Processing Systems, December 8-14, 2019, Vancouver, BC, Canada. New York: ACM, 2019: 3765-3775.
[15]	MCKENNA R, SHELDON D, MIKLAU G. Graphical-model Based Estimation and Inference for Differential Privacy[C]//PMLR. 36th International Conference on Machine Learning, June 9-15, 2019, Long Beach, California, USA. New York: ACM, 2019: 4435-4444.
[16]	ZHU Hong, ZUO Xin, XIE Meiyi. DP-FT: A Differential Privacy Graph Generation with Field Theory for Social Network Data Release[J]. IEEE Access, 2019, 19(7):304-319.
[17]	ZHU Ligeng, LIU Zhijian, HAN Song. Deep Leakage from Gradients[C]//MIT. 32nd Advances in Neural Information Processing Systems, December 8-14, 2019, Vancouver, BC, Canada. New York: ACM, 2019: 14747-14756.