Facial Recognition Tokenization via Zero-Knowledge Proof Enclaves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biometric systems, particularly facial recognition, pose significant privacy threats as they store images or models that can be used to recreate or re-generate user features, leading to potential misuse.
Innovation Solution
The implementation of zero-knowledge proof protocols and privacy-enhanced computational approaches that store signed credentials on local devices, avoid storing original data, and use targeted cryptographic mechanisms to reduce computational burden and enhance privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial recognition systems store images or models for identification, then recognition accuracy is improved, but privacy security deteriorates
Solution Approach 1:
The patent extracts and removes the original biometric images and full models from storage after credential issuance. Only transformed, privacy-preserving representations (credentials) are retained, eliminating the source material that could enable privacy breaches while preserving recognition functionality through the issued credentials.
Solution Approach 2:
The patent transforms biometric data through cryptographic operations and model transformations, changing the parameter representation from original images to signed credentials. This transformation maintains the essential identification properties while eliminating the ability to recreate original biometric data, thus resolving the contradiction between accuracy and privacy.
2Reliability
If original biometric data is stored for future verification, then verification reliability is improved, but computational security deteriorates
Solution Approach 1:
The patent segments the biometric verification process into credential issuance and credential verification phases. The original data is used only during issuance to create credentials, then discarded. Verification relies solely on the issued credentials, separating the storage requirements from the original biometric data while maintaining verification reliability.
Solution Approach 2:
The patent performs preliminary transformation of biometric data into signed credentials during the issuance phase before any verification occurs. This preliminary action converts the data into a form that maintains verification reliability while inherently preventing model leakage, as the transformed credentials cannot be used to recreate original biometric information.
3Ease of operation
If biometric models are stored on centralized servers, then system accessibility is improved, but privacy protection deteriorates
Solution Approach 1:
The patent extracts and removes sensitive biometric data from centralized storage after credential issuance. The system transitions to storing only privacy-preserving credentials on users' local devices, eliminating the centralized repository of vulnerable biometric data while maintaining system accessibility through distributed credential verification.
Solution Approach 2:
The patent introduces signed credentials as an intermediary between the user's biometric data and the verification system. These credentials serve as a mediator that enables verification functionality while preventing direct access to original biometric data, thus resolving the contradiction between accessibility and privacy protection.
Data Source
AI summary
An approach for increasing security of biometric templates is described. An improved system is adapted to split a full set of features or representations of a trained model into a first partial template and a second partial template, the second partial template being stored on a secure enclave accessible only through zero-knowledge proof based protocols. During verification using the template, a new full set of features is received for comparison, and a model is loaded based on the available portions of the model. Comparison utilizing the second partial template requires the computation of zero-knowledge proofs as direct access to the underlying second partial template is prohibited by the secure enclave.