Protection of Neural Network Models from Privacy Breach Threats in Federated Learning Using Optimization Methods