Privacy-Secured Biometric Template Generation via Irreversible Hashing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric systems face challenges in transforming size-variant biometric templates into fixed-size, privacy-secured templates while maintaining accurate matching capabilities, as previous methods either compromise security or accuracy.
Innovation Solution
The system employs irreversible lossy transformations, such as the Evergreen hash transformation, to generate transformed, privacy-secured (TPS) templates that cannot be reversed to obtain original biometric data, ensuring security and accuracy through a multi-step process involving cohort vector generation and normalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If lossless transformation techniques are applied to original biometric templates to create secondary templates, then matching accuracy is improved, but security deteriorates because the secondary template contains all information from the original template and can be reversed to reconstruct the original
Solution Approach 1:
The patent applies irreversible parameter changes by using one-way hash functions and irreversible transformations on biometric templates. The original biometric template is transformed through multiple irreversible steps including hashing, salting, and iterative transformations that change the fundamental parameters of the data while maintaining matching capability through controlled perturbation and quantization.
Solution Approach 2:
The patent introduces intermediary elements including salt values, transformation keys, and intermediate hash representations that mediate between the original biometric template and the final transformed template. These intermediaries add layers of obfuscation and prevent direct reversal while preserving the essential matching information through controlled transformations.
2Reliability
If lossy transformation techniques are applied to improve security, then security is improved, but matching accuracy deteriorates with intolerably high accuracy losses
Solution Approach 1:
The patent applies partial lossy transformation by selectively applying irreversible transformations to only certain portions or aspects of the biometric template while preserving other critical features. The transformation is applied iteratively with controlled perturbation to achieve sufficient security obfuscation while maintaining adequate matching accuracy through selective preservation of essential characteristics.
Solution Approach 2:
The patent uses controlled parameter changes including quantization, perturbation, and hashing that transform the biometric data into a different parameter space where security is improved through irreversible transformation while matching accuracy is preserved through maintained geometric or algebraic relationships in the transformed space.
3Productivity
If variable-length biometric templates are transformed into fixed-size templates, then template standardization and efficiency are improved, but matching accuracy deteriorates due to significant information loss in the translation process
Solution Approach 1:
The patent extracts essential features and characteristics from variable-length biometric templates and represents them in a fixed-size format. Rather than simply truncating or summarizing, the extraction process identifies and preserves the most critical discriminative features while discarding redundant information, maintaining matching accuracy through selective feature extraction combined with irreversible transformation.
Solution Approach 2:
The patent transforms variable-length templates into fixed-size representations by changing the parameter space through hashing and irreversible transformations. The transformation process maps variable-dimensional biometric data into a fixed-dimensional space while preserving essential matching information through controlled perturbation and quantization techniques.
Data Source
AI summary
A biometric enrollment system can include a processor and a data store that stores one or more TPS templates and a sample population. The processor can be configured to receive a reference template associated with a subject, generate a cohort set based on a plurality of templates from the sample population, and perform a one-to-many comparison process on the reference template and the cohort set, wherein an output of the one-to-many comparison process comprises one or more cohort vectors. The processor can normalize the one or more cohort vectors to obtain one or more normalized cohort vectors, and can apply a lossy transformation to the one or more normalized cohort vectors. An output of applying the lossy transformation can comprise one or more transformed, privacy-secured (TPS) templates that are stored in the database.