Biometric Template Protection (BTP) is an important component in biometric systems, aimed at protecting sensitive biometric data from being compromised. However, BTP can often be a complex topic difficult to understand for beginners. This repository provides a basic and simplified implementation of a BTP scheme for educational purposes.

My goal is to demonstrate the fundamental concepts of template protection using a simple cryptographic technique, rather than creating a production ready system designed for secure deployment.

This simplified scheme consists of two main phases: Enrollment and Authentication.

1. The biometric data (represented as an embedding embedding1) is first quantized into a binary vector (bit_vector1).
2. A secret_key is generated, which is a random binary vector of the same length as bit_vector1.
3. The client then encrypts the binarized template by XORing bit_vector1 with the secret_key to create cipher1.
4. The client stores the secret_key and sends cipher1 to the server for later authentication.

1. During authentication, a new biometric sample (embedding2) is captured and quantized to bit_vector2.
2. The client XORs bit_vector2 with the stored secret_key to create cipher2.
3. The server calculates the Hamming distance between cipher1 (from enrollment) and cipher2 (from authentication).
4. If the Hamming distance is below a certain threshold, the user is authenticated.

• Data Leakage: While the exact bits of the cipher cannot be directly determined without the secret key, the server is able to see which authentication bits differ from the enrollment, which may in time reveal details about the vector's entropy. Particularly, the server can use the differences across several users to determine which bits are reliable, similar to an Attack via Record Multiplicity. Cross-user weaknesses like this can be mitigated by performing a client-side random permutation on the enrollment vector for each user, and storing the permute order for future authentications.

• Full Access Adversary: If the adversary has access to both the client and server, they will be able to decrypt the original enrollment bit_vector by performing XOR on the client's secret_key and the server's cipher.

1. Clone the repository:

   git clone https://github.com/templateprotection/basic-btp.git
   cd btp-scheme

2. Install the required dependencies:

   pip install -r requirements.txt

3. Run the code:

   python btp.py