How to build a revocable face biometric template?

Building a revocable face biometric template involves designing a system where facial biometric data is transformed into a non-sensitive, revocable representation (e.g., a template) that can be canceled and reissued if compromised, without exposing the original face image. The core idea is to use biometric cryptosystems or feature transformations to generate a revocable template that is distinct from the raw biometric data.

Key Concepts:

1. Biometric Cryptosystems: These combine cryptography with biometrics. A cryptographic key is generated or bound to the biometric data, and the template is stored in a way that it can be revoked.
2. Feature Transformations: Techniques like cancellable biometrics transform the original biometric features into a new domain using transformations (e.g., random projections, salting, or cryptographic hashing). These transformations are irreversible and can be changed if the template is compromised.

---

Steps to Build a Revocable Face Biometric Template:

1. Preprocess the Face Image

• Normalize the input face image (e.g., resize, align, and enhance).
• Extract facial features using a face recognition model (e.g., deep learning-based feature extractors like FaceNet, VGGFace, or ArcFace).

2. Generate a Revocable Template

• Option 1: Cancellable Biometrics

  • Apply a non-invertible transformation (e.g., random projection, salting, or cryptographic hashing) to the extracted facial features.
  • For example, you can add a random key (salt) to the feature vector or use a one-way hash function to generate the template.
  • The transformation ensures that the original biometric data cannot be reconstructed from the template.
  • To make it revocable, associate a unique transformation key or token with each template. If the template is compromised, you can simply revoke it and generate a new one using a different key.

• Option 2: Biometric Cryptosystems

  • Use a cryptographic key-binding or key-generation approach.
  • For example, in a key-binding system, the facial features are combined with a cryptographic key to generate a template. The key can only be retrieved if the correct biometric data is presented.
  • In a key-generation system, the biometric data is used to generate a cryptographic key directly.
  • To make it revocable, store the template in a way that it can be canceled (e.g., using a unique identifier or token) and reissued with a new key or transformation.

3. Store the Template Securely

• Store the revocable template in a secure database or a Hardware Security Module (HSM).
• Ensure that only the transformed template is stored, not the original face image or raw features.

4. Matching and Verification

• When a user attempts to authenticate, preprocess their face image and generate a template using the same transformation or cryptographic process.
• Compare the newly generated template with the stored template using a similarity metric (e.g., Euclidean distance, cosine similarity, or Hamming distance).
• If the similarity score exceeds a predefined threshold, the user is authenticated.

5. Revocation and Reissuance

• If the template is compromised (e.g., stolen or leaked), revoke the associated token or key.
• Generate a new template using a different transformation key, random salt, or cryptographic key.
• Update the secure database with the new revocable template.

---

Example:

Imagine a system where users enroll their face biometrics for secure login:

1. During enrollment, the user’s face image is processed to extract features using a deep learning model.
2. A random salt is generated and combined with the features. The salted features are then hashed to create a revocable template.
3. The hashed template and the associated salt (or transformation key) are stored in the database.
4. During authentication, the user’s face is scanned again, and the same process (feature extraction + salt + hashing) is applied to generate a new template.
5. The system compares the new template with the stored one. If they match, access is granted.
6. If the template is compromised, the system revokes the old template and generates a new one using a different salt or key.

---

Recommended Tencent Cloud Services:

For implementing such a system, Tencent Cloud provides several services that can help:

1. Tencent Cloud Face Recognition: For extracting facial features from images.
2. Tencent Cloud Key Management Service (KMS): For securely managing cryptographic keys used in the revocable template generation process.
3. Tencent Cloud Object Storage (COS): For securely storing transformed templates and associated metadata.
4. Tencent Cloud Database (e.g., TencentDB): For storing user information, transformation keys, and revocation status.
5. Tencent Cloud Security Products: For protecting the system against unauthorized access and ensuring data privacy.

These services can be integrated to build a secure, scalable, and revocable face biometric system.