A model distributed training method based on blockchain node trust management
By utilizing node-local training behavior data to determine trust values and manage trust during distributed training, the problem of insufficient global model reliability in existing technologies is solved, thereby improving the security of distributed training and the quality of the global model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HARBIN INSTITUTE OF TECHNOLOGY (SHENZHEN) (INSTITUTE OF SCIENCE AND TECHNOLOGY INNOVATION HARBIN INSTITUTE OF TECHNOLOGY SHENZHEN)
- Filing Date
- 2024-12-26
- Publication Date
- 2026-07-24
AI Technical Summary
Existing distributed training methods for models lack reliability when aggregating global models. This is mainly because the trust management mechanism is limited to the blockchain and does not consider local training behavior data of nodes, resulting in a high risk of malicious nodes and data tampering, which affects the training quality of the global model.
By using the local model training behavior data of each computing node to determine the trust value during the distributed training process, and aggregating the global model based on trust management, including node registration, trust value calculation, penalty mechanism and global model generation, the immutability and distributed storage of blockchain are used to ensure data integrity and security.
It improves the security and reliability of distributed training computing nodes, enhances the training quality of the global model, prevents malicious node attacks, and ensures the transparency and integrity of data.
Smart Images

Figure CN120012168B_ABST