Next-Generation Human Verification & Liveness Detection
Built for the Google Solution Challenge 2026
Built with Google technologies — Flutter, Cloud Run, Vertex AI (Gemini 2.5 Flash-Lite), ML Kit, Cloud KMS, and Firestore.
Argus is an advanced identity verification platform that ensures the person on the other side of the screen is a real, living human being. By combining facial biomechanics, heart-rate estimation (rPPG), and a Google-native cryptographic ledger, Argus defends against deepfakes, replay attacks, and automated bots.
Argus doesn't just look at a photo. It actively analyzes biological signals:
- Facial Biometrics: High-speed face tracking using Google ML Kit.
- rPPG Signal Extraction: Detects the microscopic color changes in human skin caused by heartbeat pulses (photoplethysmography).
- Behavioral Intelligence: Analyzes natural blinking patterns and smooth head movements.
To stop video recordings or deepfakes, Argus issues randomized, timed challenges:
- Dynamic Instructions: Users are prompted to "Blink twice" or "Turn your head".
- Precision Timing: Tracks reactions down to the millisecond to distinguish natural human reflexes from instant, automated bot scripts.
Every successful verification is anchored to a Cryptographic Ledger powered exclusively by Google Cloud.
- Immutable Chain: Each verification record is mathematically linked to the previous one, creating an unbreakable chain of trust.
- Official Verification: Results are digitally signed by hardware-secured keys (Google Cloud KMS), making them impossible to forge.
To elevate simple rule-based detection into true "Build with AI" intelligence, Argus includes an Adaptive Multi-Modal Liveness Reasoning Agent powered by Google Gemini 2.5 Flash-Lite.
- How it works:
- On-device: Only lightweight raw signal capture (green channel time-series + behavioral timestamps) — minimal load, works on every phone (including low-end devices).
- Backend: Existing rPPG + FFT processing + behavioral data is sent as compact structured JSON.
- Gemini 2.5 Flash-Lite acts as a forensic expert: Analyzes natural heart-rate variability (HRV), evaluates realistic blink dynamics, and detects unnatural consistency typical of deepfakes or bots. It uses thinking mode for step-by-step reasoning.
- Final Score: The liveness score blends traditional math signals with intelligent AI reasoning.
- Frontend: Flutter (Dart) + CameraX / ML Kit (minimal on-device processing for low device load)
- Backend: Spring Boot 3.5 (Java 21) on Google Cloud Run (serverless, auto-scaling)
- AI Layer: Gemini 2.5 Flash-Lite via Vertex AI (Adaptive Reasoning Agent with thinking mode)
- Signal Processing: rPPG via green-channel FFT + behavioral timing analysis
- Trust Layer: Google Cloud KMS (asymmetric signing) + Firestore (immutable cryptographic ledger chain)
- Deployment: Fully serverless on Google Cloud (us-central1)
Argus tackles the rising threat of AI-generated deepfakes, replay attacks, and identity fraud that undermines trust in digital services.
Argus directly contributes to two key UN SDGs:
- SDG 9: Industry, Innovation and Infrastructure: We build resilient, inclusive, and innovative digital infrastructure. By providing affordable, secure verification, Argus enables small businesses and fintech platforms to adopt trustworthy identity systems.
- SDG 16: Peace, Justice and Strong Institutions: We strengthen institutions by reducing identity theft and fraud. This supports Target 16.9 — “Provide legal identity for all” — by making secure digital identity verification accessible and reliable.
In India, digital identity fraud is a growing challenge affecting banking (UPI, Aadhaar-linked services), online education, and government schemes (DBT - Direct Benefit Transfer).
Argus aligns with India’s SDG commitments under NITI Aayog:
- Financial Inclusion: Designed to be lightweight and work on low-end smartphones common across India, enabling secure digital access for millions in Tier-2/3 cities and rural areas.
- Fraud Protection: Protects users from rising deepfake threats in Aadhaar, banking, and government service delivery.
User-Centric Iteration: Tested with 10+ users across different devices and lighting. We incorporated feedback to improve forehead ROI detection and added AI Reasoning Explanations so users understand their verification status.
- AI Cost: ~$12–20 for 10,000 liveness verifications (using Gemini 2.5 Flash-Lite + aggressive prompt optimization).
- Infrastructure: Google Cloud Run + Firestore + KMS stays mostly within free tier for moderate usage.
- Scalability: Fully serverless architecture designed to handle thousands of concurrent verifications at minimum cost.
- Edge Processing: Facial analysis is performed locally on your device—your video feed is never saved or transmitted.
- Secure Cloud Architecture: Verification metadata is stored using enterprise-grade encryption on Google Cloud Run and Firestore.
- 100% Serverless: High availability and scaling powered by Google Cloud Run.
- Multi-language challenge support
- Advanced deepfake artifact detection using multimodal capabilities
- Integration with banking/fintech APIs for real-world deployment
This project is built for the Google Solution Challenge 2026.