Improving Fairness in Graph Neural Networks via Mitigating Sensitive Attribute Leakage

Published in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022

Recommended citation: In this work, we observe that the message-passing in GNNs would enhance the sensitive correlation and cause the sensitive leakage. Correspondingly, we propose to adaptively mask feature channels to enhance Fairness of GNNs. /files/pdf/research/SensCorr.pdf