Guardian Angel: AI-Powered Scam Protection System

🦉 Guardian Angel is an AI-powered scam protection system designed to protect senior citizens from digital fraud and phone scams using advanced multi-agent workflows.

🌟 Overview

Guardian Angel uses specialized AI agents working together to analyze and detect scam attempts across two main channels:

• Audio calls - Recorded audio transcription and scam pattern detection
• Screenshots/Images - Visual content analysis for phishing and fraud

🚀 Features

📞 Audio Scam Detection

• Recorded audio transcription using OpenAI Whisper
• 4-agent pipeline coordinated scam analysis
• Senior-friendly interface with clear threat indicators
• Automated alerts to family and authorities
• Multi-language support for global accessibility

📸 Image Scam Detection

• OCR text extraction from screenshots and images
• 7-agent coordinated analysis using AutoGen framework:
  • OCR Specialist - Text extraction
  • Link Checker - URL safety verification
  • Content Analyst - Scam pattern recognition
  • Decision Maker - Risk assessment
  • Summary Agent - Result compilation
  • Language Translation - Multi-language support
  • Data Storage - Analysis persistence
• Visual fraud detection for phishing emails, fake banking alerts, and malicious pop-ups

📁 Project Structure

Guardian_Angel/
├── audio-scam-detection/          # Audio call scam detection
│   ├── agents_audio.py           # Audio analysis agents
│   ├── app.py                    # Gradio web interface
│   ├── requirements.txt          # Python dependencies
│   ├── config/                   # Agent configurations
│   ├── tools/                    # Audio-specific tools
│   │   ├── speech_transcriber.py # Audio transcription
│   │   ├── scam_detector.py      # Scam pattern detection
│   │   ├── alert_system.py       # Family/authority alerts
│   │   └── db_connector.py       # Database operations
│   └── TrainingData/             # Training datasets
│
└── image-scam-detection/          # Screenshot/image analysis
    ├── agents.py                 # Multi-agent pipeline
    ├── app.py                    # Gradio web interface
    ├── evaluation.py             # Performance evaluation
    ├── requirements.txt          # Python dependencies
    ├── config/                   # Agent configurations
    ├── tools/                    # Image-specific tools
    │   ├── image_ocr.py          # OCR text extraction
    │   ├── url_checker.py        # URL safety verification
    │   └── formatter.py          # Output formatting
    ├── evals/                    # Evaluation datasets
    └── examples/                 # Sample test images

🛠️ Technology Stack

• AI Framework: AutoGen 0.4.0 for multi-agent orchestration
• LLM Integration: OpenAI GPT models via Gemini API endpoint
• Audio Processing: OpenAI Whisper for transcription
• Image Processing: Gemini Vision for OCR and visual analysis
• Web Interface: Gradio for senior-friendly UI
• Data Storage: MongoDB for analysis persistence
• Configuration: YAML-based agent setup

⚡ Quick Start

Prerequisites

• Python 3.8+
• Gemini API key (set as GEMINI_API_KEY)
• Required Python packages

Audio Scam Detection

cd audio-scam-detection
pip install -r requirements.txt
python app.py

Access at: http://localhost:7861

Image Scam Detection

cd image-scam-detection
pip install -r requirements.txt  
python app.py

Access at: http://localhost:7860

🎯 Use Cases

• Senior Protection: Easy-to-use interfaces for elderly users to verify suspicious calls and messages
• Recorded Analysis: Upload and analyze suspicious audio recordings and visual content
• Family Safety: Automated alerts to family members when high-risk scams are detected
• Multi-language Support: Analysis in user's native language for better comprehension
• Educational Tool: Visual demonstration of AI agent workflows for scam detection

🔧 Configuration

Both modules use YAML configuration files in their respective config/ directories:

• Agent prompts and behavior
• Detection sensitivity thresholds
• Alert escalation rules
• Model parameters and endpoints

📊 Evaluation & Testing

The image detection module includes comprehensive evaluation tools:

• Category-based testing: Authority, commerce, financial, and opportunity scams
• Performance metrics: Accuracy, precision, recall across scam types
• False positive analysis: Legitimate content verification
• Multi-language testing: Evaluation across different languages

🤝 Contributing

Guardian Angel is designed to protect vulnerable populations from sophisticated scam attempts. Contributions that improve detection accuracy, enhance user experience, or expand language support are welcome.

👥 Collaborators

• muhammadnavas
• abdulreha

---

🔒 Guardian Angel — Protecting Senior Citizens from Digital Fraud