SilentWitness

On-device AI that documents abuse without internet, protecting survivors' privacy

Overview

SilentWitness is an offline-first AI application that helps abuse survivors safely document evidence without requiring internet connectivity. Built with Gemma 4 for the Kaggle Gemma 4 Good Hackathon (Safety & Trust track).

Features

• Voice Documentation: Record evidence through voice, processed locally
• Evidence Classification: AI-powered categorization of abuse type, severity, evidence strength
• Offline-First: Works without internet - critical for survivors in unsafe situations
• Encrypted Storage: AES-256 encrypted local storage, tamper-proof
• Decoy Mode: Disguised as calculator/notes app for safety
• Legal Export: Generate court-admissible documentation format

Technical Stack

Component	Technology
AI Engine	Gemma 4 E2B (quantized)
Runtime	Ollama / llama.cpp
ML Classifier	PyTorch + ONNX
Storage	SQLite + SQLCipher
Platform	React Native + Expo

Architecture

Voice Input → Gemma 4 (Local) → Evidence Classifier → Encrypted Storage → Legal Export

Installation

# Clone repository
git clone https://github.com/YOUR_USERNAME/SilentWitness.git

# Install dependencies
pip install -r requirements.txt

# Pull Gemma model
ollama pull gemma3:1b  # or gemma4:e2b when available

# Run locally
python src/main.py

Development Status

• Project structure initialized
• Gemma 4 setup
• Voice processing pipeline
• Evidence classifier
• Encrypted storage
• Mobile app UI
• Demo video

License

MIT License - Open source for transparency and trust

Impact

• Global: 1 in 3 women experience domestic violence (WHO)
• US: 10 million abuse incidents annually
• Documentation gap: 70% of survivors lack documented evidence
• Tech gap: No offline-first abuse documentation tool exists

---

Built with Gemma 4 for the Gemma 4 Good Hackathon