Differentially Private SGDA for Minimax Problems