Federated learning is a mechanism for coordinating multiple participants to train models in response to the development of big data and artificial intelligence technologies. Federated learning not only introduces a large number of parameter exchange processes, but is also subject to attacks involving untrustworthy devices, so stronger privacy means are urgently needed to protect the data held by the parties. This paper briefly introduces the architecture and types of federated learning, analyzes the privacy risks and attack strategies, and summarizes the privacy protection technologies. By applying these technologies, this paper proposes a privacy protection scheme from three levels: center, local, local & central.