FaceID Liveness & Identity System (RGB Only)

A professional-grade, software-based face recognition system with multi-modal liveness detection. This system is designed to provide secure biometric authentication using standard webcams, eliminating the need for expensive IR or depth-sensing hardware.

---

🏗️ Directory Structure

FaceID/
├── data/                       # Local identity storage
│   └── users.json              # Encrypted face signatures
├── src/
│   ├── capture/
│   │   └── video_stream.py     # Threaded low-latency capture
│   ├── detection/
│   │   └── face_detector.py    # MediaPipe Face Mesh integration
│   ├── liveness/
│   │   ├── active/
│   │   │   └── challenge.py    # Randomized behavioral controller
│   │   └── passive/
│   │       └── texture_analysis.py # FFT & Laplacian frequency monitor
│   ├── recognition/
│   │   └── face_recognizer.py  # Geometric signature & pose-aware matching
│   └── utils/
│       ├── database.py         # JSON-based persistent storage
│       └── geometry.py         # SolvePnP & Head Pose estimation
├── main.py                     # Application entry point & State Machine
├── requirements.txt            # Dependency manifest
└── README.md                   # System documentation

---

🛠️ Technical Working Principle

The system employs a Three-Layer Security Pipeline to verify identity and ensure "proof of life."

1. Passive Liveness (Continuous Monitoring)

• Texture Analysis: Uses Laplacian Variance to calculate image sharpness, rejecting blurred low-res photos.
• Frequency Analysis (FFT): Performs a Fast Fourier Transform on the face ROI. It identifies high-frequency Moire patterns and "digital grids" typical of smartphone screens or tablets, which are invisible to the naked eye but distinct in the frequency domain.

2. Active Liveness (Behavioral Challenge)

• Challenge-Response: Randomly requests actions (Blinks, Head Turns).
• Relative Pose Tracking: Instead of checking for absolute angles, the system captures the user's initial orientation and measures the angular delta. This prevents static photos or "tilting images" from passing, as it requires true 3D volumetric rotation.

3. Identity matching (Geometric Signature)

• Landmark Encoding: Extracts 40+ high-precision landmarks.
• Normalization: Landmark coordinates are normalized relative to the nose tip and scaled by the inter-ocular distance (face width), making the signature invariant to distance from the camera.
• Pose-Aware Buffering: During challenges, the system buffers face signatures and automatically selects the most front-facing sample for comparison against the database.

---

🚀 Getting Started

Prerequisites

• Python 3.10+
• A standard RGB Webcam

Installation

1. Clone & Navigate:

   cd FaceID

2. Install Dependencies:

   pip install -r requirements.txt

Operational Workflow

1. Launch: Run python main.py.
2. Register ('r'): Enter your name in the terminal. The system will run a liveness check. Once passed, your geometric signature is saved to users.json.
3. Login ('l'): The system will verify your liveness first. If successful, it will scan the database and authenticate your identity.

---

📊 Security Benchmarks

Threat	Defense Layer	Mitigation Strategy
Printed Photo	Active Liveness	Fails EAR (Blink) and SolvePnP (3D Pose).
Video Replay	Passive Liveness	FFT detects Moire patterns and LCD refresh artifacts.
3D Mask	Passive Liveness	Frequency analysis detects unnatural skin texture & lack of pore-level detail.
Static Image Tilt	Relative Pose	Fails to produce the correct 3D geometric landmark shift.

---

📄 License

This project is licensed under the MIT License. Developed for research in biometric security and agentic automation.