Encrypted Neural Network Weights for Biometric Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods fail to encrypt biometric information effectively for authentication without degrading recognition performance, leading to security vulnerabilities and unacceptable trade-offs in memory and speed requirements.
Innovation Solution
The implementation of a system that uses homomorphic encryption and multi-party computation to store and process encrypted biometric data, allowing for secure authentication with undegraded recognition performance by encrypting trained neural network weights and performing computations on encrypted data, while ensuring only the user with the private key can decrypt the match-score for authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If biometric information is encrypted using conventional methods, then security is improved, but recognition performance degrades
Solution Approach 1:
The system performs preliminary encryption of the trained neural network weights during the enrollment phase. The encrypted weights are stored and can be used directly during authentication without requiring decryption, thus maintaining both security and recognition performance. This preliminary action eliminates the need to decrypt sensitive data during authentication operations.
Solution Approach 2:
The patent introduces an intermediary cryptographic transformation that allows computations to be performed on encrypted data. The encrypted neural network weights act as an intermediary representation that preserves the functional capabilities of the original weights while maintaining security. This intermediary form enables the system to operate on biometric data without exposing the actual biometric templates.
2Ease of operation
If biometric information is stored in databases, then authentication functionality is maintained, but vulnerability to attacks increases
Solution Approach 1:
The system extracts the sensitive biometric information from the authentication process by encrypting the neural network weights. The encrypted weights are stored in a database, but the actual biometric data is never stored in plaintext. This extraction of sensitivity allows the system to maintain authentication functionality while removing the vulnerability associated with storing raw biometric data.
Solution Approach 2:
The patent transforms the biometric data through cryptographic parameter changes. The original biometric data is transformed into encrypted weights through a cryptographic function, changing the parameter representation from human-interpretable biometric features to encrypted numerical values. This parameter change maintains the functional properties needed for authentication while making the data resistant to attacks.
3Reliability
If encryption methods are applied to biometric data, then security claims are strengthened, but memory and speed requirements increase
Solution Approach 1:
The encryption of neural network weights is performed preliminarily during enrollment, and the encrypted form is stored for reuse. This eliminates the need to perform computationally intensive encryption operations during each authentication, thus reducing the memory and speed requirements during actual authentication while maintaining strong security claims.
Data Source
AI summary
Embodiments of the present systems and methods may provide encrypted biometric information that can be stored and used for authentication with undegraded recognition performance. For example, in an embodiment, a method may comprise storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature, obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, and authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.