Tag
This paper introduces a mean-field privacy game framework for federated learning, enabling tractable Nash equilibrium analysis for arbitrarily many clients with heterogeneous privacy preferences and yielding a personalized privacy guarantee.