A lateral federated learning (ABC-2RDP-FL) privacy protection method based on dual Rényi differential privacy and artificial bee colony is proposed to address the poor performance of federated learning in balancing privacy protection and model utility. In ABC-2RDP-FL, a dual Rényi differential privacy protection mechanism was designed to more tightly measure privacy budget and improve privacy protection performance. Then, a federated learning hyperparameter optimization algorithm based on artificial bee colonies was proposed to balance privacy protection and model utility.