backend/dataset/metadata.json

{
    "version": "1.0",
    "name": "Indian Multi-Modal Scam Dataset",
    "description": "Curated dataset of India-specific scam patterns covering UPI fraud, KYC phishing, lottery schemes, OTP theft, WhatsApp forwards, job fraud, and phishing URLs. Used to enhance zero-shot classification and keyword-based scoring in the ScamDetect AI pipeline.",
    "categories": [
        "upi_fraud",
        "kyc_phishing",
        "lottery_scam",
        "otp_theft",
        "whatsapp_forward",
        "job_scam",
        "investment_scam",
        "impersonation",
        "delivery_scam",
        "tax_scam",
        "url_phishing",
        "benign"
    ],
    "modalities": [
        "text",
        "url",
        "image_ocr"
    ],
    "statistics": {
        "total_samples": 50000,
        "text_messages": 30000,
        "urls": 20000,
        "images": 5000,
        "videos": 2000
    }
}