Feature 9: AI Fraud Detection for Scans & Images

[saidai-bhuvanesh]

Description

To prevent users from uploading fake photos, edited images, or duplicate screenshots to manipulate vendor ratings, this issue proposes an AI Fraud Detection module.

We will build a detector on the backend analyzing image statistics, and check for artificial color saturation edits (e.g. painting gills redder). The API will return warning flags, and the UI will alert users and flag suspicious scans.

Technical Implementation Details

1. Tampering Detection: Detect artificial color enhancement by comparing red-channel color margins.
2. Duplicate Check: Match duplicate scans to block trust score manipulation.
3. UI Warning Banner: Display a warning banner on the dashboard if fraud is suspected.

[arcgod-design]

Hi @jpdevhub! I'd love to build this as part of SSoC 2026 👋

My Approach:

• Train/integrate a classifier to detect common fish scan fraud: image manipulation (Photoshop artifacts, splice detection), metadata inconsistencies, and repeated/duplicate image submissions
• Use perceptual hashing (pHash) to detect near-identical images submitted multiple times with different results
• Add EXIF metadata validation: flag images with suspicious editing timestamps or missing geolocation data
• Display a "Fraud Risk" indicator in scan results with explanation

Could you please assign this to me (@arcgod-design)? Thank you!