2.3.4 Decentralized Distributed Storage of Encrypted Data;

The node client of BOBchain block network will be integrated with IPFS by the node client of BOINC AI server for storing and transferring large model files, datasets. The model files are encrypted and split into smaller chunks and transmitted using multiple paths. This increases the transmission speed and reduces the impact of single path failures on the overall transmission. Encrypt large files in segments and upload them to different nodes and transmit different chunks in parallel using multi-path. This can further increase the transmission speed. Adopt P2P intranet penetration technology, so that the intranet nodes are directly interconnected and do not need to transit through the public network. This can improve transmission efficiency and reduce transmission delay.

BOBchain network uses cryptographic zero-knowledge proofs for all data such as model data, datasets, user information, etc. in AI model training. Transmission of data encrypted in transit using Transport Layer Security (TLS) protocol. Encrypted storage is distributed across multiple nodes. Only the content path hash is recorded on the chain technical solution to ensure privacy security and data security. The hash chain ensures data integrity and consistency. Each file has a unique hash value, providing data tamper-proof protection for AI training.

The protection of the interests of developers of modelling techniques and data providers is further ensured.

In addition, in the storage of model training data.BOBchain uses lossless or accuracy-controlled lossy compression algorithms such as fixed-point compression, pruning, and quantization to reduce the file size. This reduces the amount of data transferred. Periodically, the old and new models are compared. Only the difference part of the update is transmitted. This can further reduce the amount of data transferred. Remove redundant and non-critical parameters to reduce model size. Trim layers and parameters that are used for training but not for inference, and keep only the sub-networks that are useful for inference. This can further reduce the model size.

With these technologies, BOBchain is able to securely and efficiently store and transfer large model files, providing a powerful distributed storage solution for AI model training and reducing reliance on hard drives. This helps improve the performance and efficiency of AI blockchain networks for applications such as AI training.