With the development of attack methods in the information age, the traditional rule-based network access and security assessment can hardly meet the demand. This paper focuses on the design and implementation of a network access security assessment system, which uses trustworthy federated learning technology to solve the problem of multi-participant collaborative training of security state detection models. The system uses the trustworthy federated learning module to determine the piecewise aggregation scheme before model aggregation and identifies and removes malicious participants during the aggregation process. The robustness and adaptability of the system are improved through layered and modularized design. Experimental results show that the designed system is capable of continuous access authentication and security state assessment of network devices in energy Internet scenarios, and collaboratively trains network state assessment models using the trustworthy federated training process.