Encrypted Machine Learning Classification Without Data Decryption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical decision-making systems face challenges in secure data and model sharing due to privacy concerns, leading to delayed adoption and compromised patient care quality.
Innovation Solution
A system utilizing fully homomorphic encryption techniques to enable secure classification of encrypted data using an encrypted machine learning model, allowing secure data transmission and processing without decryption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data and models are shared in unencrypted form to enable classification, then classification accuracy and system functionality are improved, but data privacy and security are compromised
Solution Approach 1:
The patent introduces fully homomorphic encryption (FHE) as an intermediary mechanism that allows computation on encrypted data without decryption. The FHE-encrypted model and data act as intermediaries that enable classification functionality while maintaining privacy, resolving the contradiction between sharing data for classification and protecting data privacy.
Solution Approach 2:
The patent changes the state of data from unencrypted to encrypted form while maintaining computational functionality through FHE properties. By transforming data into encrypted parameters that still support mathematical operations, the system achieves both privacy protection and classification capability.
2Reliability
If models are encrypted to protect intellectual property, then model security and data owner privacy are improved, but model sharing and collaboration are restricted
Solution Approach 1:
FHE serves as an intermediary that enables encrypted model sharing. The encrypted model can be deployed across multiple hospitals and organizations without exposing the underlying model parameters, allowing collaboration and adaptability while maintaining model security and intellectual property protection.
Solution Approach 2:
The encrypted model system achieves multi-functionality by enabling the same encrypted model to serve multiple data owners across different hospitals and organizations. The universal encrypted model can be shared and executed across diverse environments without requiring decryption or exposure of proprietary parameters.
3Reliability
If data is decrypted before classification to ensure processing accuracy, then classification reliability is improved, but data exposure and privacy breaches increase
Solution Approach 1:
The FHE encryption acts as an intermediary layer that maintains data in encrypted form throughout the classification process. This eliminates the need to decrypt data before processing, thereby preventing data exposure while still achieving accurate classification through homomorphic computation operations on encrypted data.
Data Source
AI summary
Systems and methods for classifying encrypted data using an encrypted machine learning model are disclosed. An example method includes receiving, at one or more processors, encrypted data from a user that is encrypted in accordance with a first fully homomorphic encryption technique. The example method further includes analyzing, by the one or more processors executing an encrypted ML model that is encrypted in accordance with a second fully homomorphic encryption technique, the encrypted data to output an encrypted classification without decrypting the encrypted data. The example method further includes transmitting, by the one or more processors, the encrypted classification to a user computing device for decryption.