docs/detector/meta.md

Metadata Analysis for Image Edit Detection

This module inspects image metadata to detect possible signs of AI-generation or post-processing edits.

Overview

• Many AI-generated images and edited images leave identifiable traces in their metadata.
• This detector scans image EXIF metadata and raw bytes for known AI generation indicators and common photo editing software signatures.
• It classifies images as "ai_generated", "edited", or "undetermined" based on detected markers.
• Handles invalid image formats gracefully by reporting errors.

How It Works

• Opens the image from raw bytes using the Python Pillow library (PIL).
• Reads EXIF metadata and specifically looks for the "Software" tag that often contains the editing app name.
• Checks for common image editors such as Photoshop, GIMP, Snapseed, etc.
• Scans the entire raw byte content of the image for embedded AI generation identifiers like "midjourney", "stable-diffusion", "openai", etc.
• Returns a status string indicating the metadata classification.

🔙 Back to Main README