Fair and Privacy-Preserving Graph Neural Network

Xuemin Wang, Tianlong Gu, Xuguang Bao, Liang Chang

Abstract: Graph neural networks (GNNs) have demonstrated superior performance in modeling graph-structured. They are vastly applied in various high-stakes scenarios such as financial analysis and social analysis. Among the fields, privacy issues and fairness issues have become the focus of attention. Currently, most studies focus on protecting data privacy or promoting the fairness of the model. However, how to both guarantee privacy and fairness is under-explored on GNN. To ensure GNNs behave in a socially responsible manner, it is necessary to protect the privacy and mitigate bias simultaneously. Therefore, our Ph.D. project aims at creating GNNs which can promote fairness and protect data privacy, preserving high utility performance.