Datalake FaceID — Offline-First Biometric Attendance Platform

An enterprise-grade, offline-first facial recognition and active liveness verification gateway. This system coordinates biometric attendance tracking on low-bandwidth edge devices in remote zero-network zones, complete with a dynamic local database fallback and a secure AWS sync-and-purge cycle.

---

🚀 Key Features & Architectural Highlights

1. Dual-Core Database Engine (MongoDB & LocalDB)

   • Local Backend Database: Powered by Express & Mongoose, storing biometric registers, logs, and synchronization checkpoints in a local MongoDB service (mongodb://localhost:27017/faceid).
   • Browser LocalStorage Fallback: Dynamically toggles to client-side localStorage if the server goes offline, allowing 100% functionality with zero network connectivity.

2. Ultra-Lightweight Edge AI Stack (< 8.0 MB Footprint)

   • Active Landmarking & Mesh: Utilizes a lightweight landmark extraction mesh (3.0 MB) mapping 468 facial points to track movements in real-time.
   • Identity Embedding Matcher: Employs a quantized MobileFaceNet engine (4.5 MB) generating 128-dimensional facial vectors. Matches registered templates locally via Cosine Similarity / L2 Euclidean Distance in ~42 ms.
   • Interactive Liveness Spoof Defense: Implements challenge-response protocols detecting Eye Blink (Eye Aspect Ratio), Smiles (Mouth Aspect Ratio), and Head Yaw/Pitch rotations to block photo and video spoofing attempts.

3. Secure AWS Sync-and-Purge Protocol

   • Retains base64 snapshots and credentials locally inside offline write-ahead logs.
   • Restoring internet connectivity triggers batched data streams uploading to AWS S3 & DynamoDB.
   • Upon receiving receipt confirmation, a secure local purge deletes cached base64 images to free disk space while retaining rolling metadata attendance reports.

---

📂 Project Structure Map

• 📂 react-native-prototype/ — Native Edge AI Prototype Codebase
  • App.tsx — Coordinator dashboard dashboard containing biometrics verification triggers.
  • components/LivenessScanner.tsx — Native camera viewport rendering scanning guides and step-by-step gesture HUD overlays.
  • services/LivenessEngine.ts — EAR/MAR and head rotation angle mathematical state machine.
  • services/FaceRecognitionEngine.ts — Offline TFLite wrapper cropping face frames and loading quantized templates.
  • services/SyncPurgeManager.ts — SQLite storage engine handling write-ahead logs, net-state listeners, and AWS push/purge queues.
• 📄 index.html — Premium, fully responsive Web Dashboard simulator showing facial landmark scanning sweeps, stats trends, and network synchronization panels.
• 📄 server.js — Local Express backend api exposing personnel database streams.
• 📄 generate_assets.py — Python pptx automation script programmatically compiling professional deck slide presentations.
• 📄 technical_documentation.md — Extensive technical documentation including mathematical equations and architecture blueprints.

---

🛠️ Quick Start Instructions

Prerequisites

• Node.js (v18+)
• MongoDB (running locally on port 27017)
• Python (optional, for asset generation)

1. Run the Local Server & Dashboard

1. Install dependencies:

   npm install

2. Start the Express backend API:

   npm run dev

3. Double-click index.html to open the biometric dashboard or access it via http://localhost:5000.

2. Generate the Hackathon Pitch Deck

To compile the premium PowerPoint presentation containing architecture benchmarks and equations:

python generate_assets.py

This generates presentation.pptx directly in the project directory.

---

🌐 Public Tunneling (Sharing your Proposal Link)

To share this running system with external reviewers or test on smartphones instantly:

1. Keep the local backend running (npm run dev).
2. Open a second terminal window and run:

   ssh -R 80:localhost:5000 nokey@localhost.run

3. Copy the generated https://xxxx.lhr.life link printed on your terminal screen and share it. The dynamic API layer built inside index.html automatically adapts and routes requests correctly!

---

🛡️ Privacy & Compliance

All biometric images are completely blurred using CSS filters in the dashboard for demographic privacy, and descriptions are stored as mathematical arrays rather than raw face images.