In the problem of privacy computing, Secure Multi-Party Computation (MPC) is an indispensable and important means. Confidentiality calculation mean value is one of the important research problems of secure multi-party computation (MPC). Its core goal is to complete the calculation of mean value collaboratively without leaking the data of multiple parties. For example, in federated learning, multiple participants need to calculate the global average of model parameters (such as gradient aggregation), but they need to protect local training data. However, most of the current computing protocols use public key encryption to protect data privacy, and the computing cost is large. When users with weak computing power solve complex problems, the computing efficiency is often low. In view of the above problems, this paper designs a new protocol, which does not need to use public key encryption. In the plaintext state, it can achieve the purpose of confidential calculation of average value by combining mathematical thinking and cryptographic tools. The protocol designed in this paper does not need to pass public key encryption, which reduces the computational complexity relatively and belongs to information theory security theory.