🛡️ DefenceX

AI-powered scam call detection — transcribe, analyse, and flag suspicious calls in 33 languages.

---

Overview

DefenceX is a web app that helps users decide whether a phone call is a scam. Upload an audio recording, pick a language, and the service transcribes it via Google Cloud Speech-to-Text and runs the transcript through Gemini for scam-pattern analysis. Results are rendered back to a simple UI.

Features

• 🎙️ Audio upload — drag-and-drop interface for call recordings
• 🌐 Multilingual transcription — 33 languages via Google Cloud Speech-to-Text
• 🔁 Format normalisation — incoming audio is converted to WAV before transcription
• 🧠 Gemini-powered analysis — transcript is scored for scam indicators
• 🖥️ Web UI — Flask backend renders findings in the browser

Tech Stack

Layer	Tools
Backend	Flask (server.py)
Speech	Google Cloud Speech-to-Text
LLM analysis	Gemini API
Frontend	HTML template (templates/index.html)

Getting Started

# 1. Clone
git clone https://github.com/bhavya-x/DefenceX.git
cd DefenceX

# 2. Virtualenv
python -m venv venv
source venv/bin/activate          # Windows: venv\Scripts\activate

# 3. Install
pip install -r requirements.txt

# 4. Configure credentials
export GOOGLE_APPLICATION_CREDENTIALS=path/to/service-account.json
export GEMINI_API_KEY=your_key_here

# 5. Run
python server.py
# Open http://localhost:5000

Project Structure

DefenceX/
├── server.py            # Flask app + transcription/analysis pipeline
├── requirements.txt     # Python dependencies
└── templates/
    └── index.html       # Upload UI + result view

Roadmap

• Live-call streaming mode
• Confidence scoring & reasons in the UI
• Caller-pattern history per user
• Self-hosted Whisper fallback for offline mode