Face Authentication with Liveness Detection

This project is a Python-based face authentication system that combines face recognition with liveness detection to prevent spoofing attacks (e.g., using photos or videos).

Features

• Face Registration & Recognition

  • Uses MTCNN for face detection.
  • Uses FaceNet (InceptionResnetV1) for generating face embeddings.
  • Stores embeddings locally in faces_db/.

• Liveness Detection (Anti-Spoofing)

  • Blink Detection – verifies natural eye blinks.
  • Head Movement – ensures the user turns left/right.
  • Smile Detection – detects mouth widening and lip movement.
  • Randomized challenges make it harder to cheat.

• Interactive Menu (OpenCV GUI)

  • Register new users.
  • Login with liveness + face recognition.
  • Quit the system.

Tech Stack

• Python, OpenCV – video capture & display.
• Mediapipe – facial landmarks (eyes, mouth, head movement).
• Facenet-PyTorch – face embeddings & recognition.
• NumPy, Torch – vector computations & ML.

Workflow

1. Register → Capture 50 samples of user face, store average embedding.
2. Login → Run randomized liveness challenges (blink, head, smile).
3. Authenticate → Compare captured embedding against saved embeddings.
4. Result → Access granted if similarity > threshold.

Usage

1. Run the Python script.
2. Use the menu to register or login.
3. Follow the on-screen prompts for liveness challenges.

Directory Structure

faces_db/      # Stores user embeddings
test.py        # Main script with all functionality