VoxGuard is a real-time cybersecurity tool designed to detect AI-generated deepfakes and social engineering scams in voice notes. By analyzing both the biometric "voiceprint" and the linguistic intent of a message, VoxGuard provides a comprehensive security report to protect users from high-tech fraud.
Built as a hackathon project, this system bridges the gap between complex AI threat detection and an intuitive, user-friendly interface.
- Biometric Audio Processing: Converts raw audio into highly accurate text using Azure's enterprise-grade Speech Services.
- Intent Analysis Engine: Utilizes Google's Gemini 2.5 Flash model to perform behavioral analysis on the transcript, identifying psychological manipulation, urgency triggers, and scam playbooks.
- Threat Scoring: Generates a quantifiable "Risk Score" (0-100%) alongside clear, actionable bullet points explaining the red flags.
- Modern UI: A sleek, Cyberpunk-themed web dashboard built with Python Streamlit for frictionless drag-and-drop analysis.
- Frontend: Streamlit (Python)
- Audio Processing (The "Ears"): Azure Cognitive Services (Speech-to-Text SDK)
- Threat Intelligence (The "Brain"): Google GenAI (
gemini-2.5-flash) - Environment Management:
python-dotenv
Follow these steps to run VoxGuard locally on your machine.
- Python 3.9+ installed.
- An Azure account with a Speech Services resource (Key & Region).
- A Google AI Studio account with a Gemini API Key.
- Clone the repository:
git clone [https://github.com/amay09x/VoxGuard.git](https://github.com/amay09x/VoxGuard.git) cd VoxGuard