FHE-NIT System for Secure Encrypted ML Training and Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning technologies face challenges in securely training and inferring insights from encrypted data, particularly in multi-party settings where data cannot be shared due to sensitivity and compliance issues, and existing solutions are inefficient for training deep neural networks using fully homomorphic encryption.
Innovation Solution
The development of a Fully Homomorphic Encryption based Network for Inferencing and Training (FHE-NIT) system that enables secure multi-party learning and inference by using fully homomorphic encryption (FHE) to train machine learning models on encrypted data, allowing for secure sharing and processing without revealing raw data, and incorporating techniques for efficient operation distribution and accuracy enhancement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fully homomorphic encryption is used to train machine learning models on encrypted data, then data privacy and security are improved, but training time and computational complexity increase significantly
Solution Approach 1:
The system performs preliminary actions by pre-processing and encrypting training data using fully homomorphic encryption before the actual model training process. This allows the encryption scheme to be established and optimized in advance, enabling efficient encrypted computations during training without sacrificing data privacy
Solution Approach 2:
The patent introduces an intermediary encrypted computation layer that mediates between the raw encrypted data and the machine learning model. This intermediary layer enables computations to be performed on encrypted data through homomorphic operations, allowing training to proceed without decryption while maintaining acceptable computational efficiency
2Reliability
If fully homomorphic encryption is used for model training, then data security is improved, but computational complexity and resource requirements increase
Solution Approach 1:
The training process is segmented into distinct phases: data encryption phase, model initialization phase, encrypted training phase, and inference phase. Each phase uses optimized computational approaches tailored to its specific requirements, reducing overall computational complexity while maintaining security
Solution Approach 2:
The system changes computational parameters by using approximate homomorphic encryption schemes that balance security with computational efficiency. It adjusts encryption parameters, precision levels, and computation depth to achieve acceptable trade-offs between data security and computational complexity
3Measurement precision
If multiple parties share encrypted data for collaborative learning, then model accuracy is improved, but system complexity and coordination overhead increase
Solution Approach 1:
The patent creates a universal encrypted data sharing framework that serves multiple functions: secure data exchange, collaborative model training, and joint inference. This multi-functional system allows multiple parties to participate in collaborative learning without requiring separate protocols for each operation, reducing overall system complexity
Data Source
AI summary
ML model(s) are created and trained using training data from user(s) to create corresponding trained ML model(s). The training data is in FHE domains, each FHE domain corresponding to an individual one of the user(s). The trained machine learning model(s) are run to perform inferencing using other data from at least one of the user(s). The running of the ML model(s) determines results. The other data is in a corresponding FHE domain of the at least one user. Using at least the results, it is determined which of the following issues is true: the results comprise objectionable material, or at least one of the trained ML model(s) performs prohibited release of information. One or more actions are taken to take to address the issue determined to be true. Methods, apparatus, and computer program product are disclosed.