Blockchain-secured AI Module Transfer and Customization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training machine learning systems with insufficient data can lead to user frustration, as they are difficult to customize for specific tasks, and existing solutions lack secure and private communication methods for AI modules across environments.
Innovation Solution
The implementation of a system that securely and privately communicates AI modules by encrypting them and adding them to a blockchain, allowing for secure transfer between devices while erasing personal information, and utilizing a customization system to train AI modules on one device and share them across multiple local devices for consistency and resource conservation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If AI modules are trained on local devices with personal information, then customization and accuracy are improved, but security and privacy are compromised when transferring modules
Solution Approach 1:
The patent extracts and removes personal information from AI modules before transfer. The system identifies and erases identifiable features such as names, addresses, and other personal data from training datasets and corresponding AI module parameters, allowing the module to be transferred without compromising privacy while maintaining functional accuracy.
Solution Approach 2:
The patent introduces an intermediary blockchain system that facilitates secure transfer of AI modules. The blockchain acts as a trusted intermediary that verifies module authenticity, tracks transfer history, and enables decentralized verification without requiring direct exposure of personal information between sender and receiver.
2Adaptability or versatility
If AI modules are trained independently on each device, then local customization is achieved, but device resources are wasted through redundant training
Solution Approach 1:
The patent merges training efforts across multiple devices by creating a shared AI module repository on the blockchain. Instead of each device training independently, devices can access pre-trained modules from the blockchain, combining computational resources and eliminating redundant training while maintaining local customization through selective module deployment.
Solution Approach 2:
The patent performs preliminary training and customization of AI modules on a centralized or selective basis before deployment to multiple devices. By pre-training modules with common functionality and storing them on the blockchain, the system performs the computationally intensive training action once rather than repeatedly, saving resources while allowing local adaptation.
3Speed
If AI modules are transferred without encryption, then transfer speed is improved, but security and privacy are compromised
Solution Approach 1:
The patent applies encryption to AI modules before transfer, preparing the data in advance for secure transmission. The encryption process is performed preliminarily, and the encrypted modules are stored on the blockchain where they remain secure during transfer, ensuring both security and efficient retrieval without requiring encryption during the actual transfer operation.
Data Source
AI summary
A user can customize (also referred to as train) an AI module, which includes any of a variety of machine learning systems. The AI module can be used by the device on which the AI module is trained or can be communicated to other devices in the same environment (e.g., the same home). The AI module can also be communicated in a secure and private manner to a device in another environment (e.g., another user's home). To do so, the AI module is encrypted and added to a blockchain, and the blockchain is communicated via a peer-to-peer network to the device in the other environment. The recipient of the blockchain can then decrypt the AI module and use the AI module in that other environment, including further training the AI module for use in that other environment.