MVP Live · Google Solution Challenge
🛡️ AthenaGuard
Technical Docs
AI-powered sports media protection. Detects piracy even after cropping, mirroring, meme edits, and compression.
TypeScript 99.7%
Next.js 14
Gemini API
Firebase
GCP
MIT License
01 / Problem Statement
Why AthenaGuard?
Sports broadcasters lose billions annually to unauthorized media redistribution. Existing copyright detection tools rely on exact matching or watermarks — both fail when content is transformed.
The Gap: Traditional systems cannot detect content that has been cropped, mirrored, meme-edited, compressed, or re-uploaded as short clips. AthenaGuard fills this gap with transformation-aware AI.
| Transformation | Traditional Tools | AthenaGuard |
|---|---|---|
| Exact match | ✓ Detected | ✓ Detected |
| Cropped / trimmed | ✗ Missed | ✓ Detected |
| Mirrored / flipped | ✗ Missed | ✓ Detected |
| Meme overlay | ✗ Missed | ✓ Detected |
| Compressed / re-encoded | ✗ Missed | ✓ Detected |
| Short clip re-upload | ✗ Missed | ✓ Detected |
02 / Solution
What AthenaGuard Does
A full-stack AI platform providing end-to-end protection: from ingestion and detection through enforcement and analytics.
Transformation-Aware Detection
Detects misuse via embeddings + pHash even after heavy edits
Real-Time Monitoring
Queue-based simulated stream scanning multiple platforms concurrently
Manual Verification
Drag-and-drop forensic upload returning similarity score + action
Automated DMCA
One-click notice generation, simulated dispatch, and status tracking
ROI Dashboard
Revenue protected, violation heatmaps, platform-wise breakdown
AI Explainability
Gemini-powered reasoning — why content was flagged, audit trail
03 / Architecture
System Architecture
Five distinct layers with clear separation of concerns:
User Layer
Admin / Analyst → Login (Firebase Auth) → Dashboard → Mode selection
↓
Ingestion Layer
Simulated stream queue ↔ Manual upload interface
↓
AI Core
Preprocessing → Frame extraction → Embeddings + pHash + OCR → Similarity matching → Decision engine
↓
Storage Layer
Media DB · Vector DB (embeddings) · Incident DB · User DB
↓
Output Layer
Dashboard · Alerts · DMCA generate → send → track · ROI analytics
Decision Engine Thresholds
| Confidence | Action |
|---|---|
| High > 90% | Auto-generate DMCA notice + dashboard alert |
| Medium 75–90% | Route to analyst manual review queue |
| Low < 75% | Log for audit, no action taken |
04 / Tech Stack
Technologies Used
Frontend
| Technology | Purpose |
|---|---|
React.js / Next.js 14 | Responsive dashboard UI + server-side rendering |
Tailwind CSS | Utility-first styling |
TypeScript | Type-safe codebase (99.7% of repo) |
AI & Machine Learning
| Technology | Purpose |
|---|---|
Google Gemini API | Content understanding, DMCA generation, AI explanations |
OpenCV | Image and video frame processing |
Embeddings | Semantic similarity detection across transformations |
pHash | Transform-resistant perceptual fingerprinting |
OCR (Tesseract / Cloud Vision) | Meme text recognition and extraction |
Cloud & Infrastructure
| Technology | Purpose |
|---|---|
Google Cloud Platform | Core infrastructure, compute, hosting |
Firebase / Firestore | Real-time database and authentication |
Cloud Storage | Media file and asset storage |
FAISS / Pinecone | Fast vector similarity search |
Firebase Auth | Secure login, role-based access (Admin / Analyst) |
05 / Getting Started
Installation
Prerequisites
Node.js 18+, npm or yarn, Google Gemini API key (aistudio.google.com), and a Firebase project.
Clone & Install
git clone https://github.com/CodeGeek-Garvit/AthenaGuard1.git cd AthenaGuard1 npm install cp .env.example .env.local
Run
npm run dev # Development → http://localhost:3000 npm run build # Production build npm start # Start production server
Environment Variables
.env.local
GEMINI_API_KEY=your_gemini_api_key_hereREQUIRED
NEXT_PUBLIC_FIREBASE_API_KEY=your_firebase_api_key
NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN=your_project.firebaseapp.com
NEXT_PUBLIC_FIREBASE_PROJECT_ID=your_project_id
NEXT_PUBLIC_FIREBASE_STORAGE_BUCKET=your_project.appspot.com
NEXT_PUBLIC_FIREBASE_MESSAGING_SENDER_ID=your_sender_id
NEXT_PUBLIC_FIREBASE_APP_ID=your_app_id
06 / Features
Key Features
AI-Based Media Detection
Dual-fingerprint approach combining deep learning embeddings with perceptual hashing ensures high recall across transformation types with controlled false-positive rates. OCR adds meme text recognition as a third detection signal.
Real-Time Monitoring (Simulated)
Queue-based streaming pipeline simulates live platform ingestion with configurable concurrent workers. Designed for drop-in replacement with YouTube Data API v3 and Instagram Graph API in the next phase.
Automated DMCA Workflow
| Status | Description |
|---|---|
Generated | Notice auto-created with evidence attachment |
Sent | Simulated dispatch to platform Trust & Safety |
Under Review | Platform acknowledgment received |
Resolved | Content removed or dispute resolved |
Dashboard Views
| View | Content |
|---|---|
| Security Overview | Revenue protection, violation count, confidence stats, monitored reach |
| Monitoring Feed | Live violation stream with transformation tags |
| Incident Analysis | Side-by-side original vs detected + AI explanation |
| DMCA Notice | Auto-generated legal notice with export & send options |
| Manual Verification | Drag-and-drop forensic scanning interface |
| Media Library | Protected asset management with violation tracking |
07 / Process Flow
End-to-End Flow
1
User authenticates via Firebase Auth and selects Monitor or Verify mode
2
Monitor: simulated stream feeds media to ingestion queue across platforms
3
Verify: analyst uploads image or video clip for on-demand forensic analysis
4
AI Detection Engine extracts features: embeddings, pHash, OCR text
5
Similarity matching runs against protected media library in vector DB
6
Decision Engine classifies: High (auto-DMCA) · Medium (review queue) · Low (log)
7
DMCA flow: Generate notice → Simulated send → Track status to resolution
8
Results surface on dashboard with ROI metrics and AI explanation audit trail
08 / Cost
Cost Estimation
Development cost: ₹0 (student-built). Production cloud estimates below.
| Service | Monthly (INR) |
|---|---|
| Google Cloud (Compute + Storage) | ₹3,000 – ₹6,000 |
| Firebase (DB + Auth) | ₹1,000 – ₹2,000 |
| Gemini API (AI usage) | ₹2,000 – ₹5,000 |
| Total | ₹6,000 – ₹13,000 |
Architecture is cloud-native and pay-as-you-go. Vector search optimization and modular microservices ensure cost scales linearly with usage.
09 / Stakeholders
Ecosystem Stakeholders
| Stakeholder | Role | Value Delivered |
|---|---|---|
| BCCI | Primary Content Owner | IPL digital rights protection |
| Star Sports | Official Broadcaster | Exclusive broadcast slot verification |
| JioCinema | Streaming Partner | Piracy-led churn reduction |
| Trust & Safety Teams | Platform Moderators | Pre-verified flags, 40% faster review |
10 / Roadmap
Development Roadmap
NOW — MVP
Current Release
- Simulated real-time monitoring
- Manual verification interface
- DMCA notice generation
NEXT — Platform Integration
Live API Connections
- YouTube Data API v3
- Instagram Graph API
- Real-time takedown workflows
v3 — Advanced AI
Enhanced Detection
- Full video-level tracking
- Live stream detection
- Improved short-clip analysis
v4 — Automation
Legal at Scale
- Direct legal service integration
- Faster DMCA execution at scale
v5 — Global
Global Expansion
- OTT platform support
- Blockchain-based ownership verification
- Tamper-proof media authenticity tracking