← Back to WorksProject Detail // 2025
Machine Learning
Screenshot Forgery Detector
Role
ML Engineer
Timeline
2025
Tech Stack
Pythonscikit-learnOpenCVDeep Learning
Case Study Overview
With the rise of digital banking, tampered and forged screenshots of payment receipts are increasingly used to commit financial fraud.
01 // Context
Project Summary
Developed an advanced forensic machine learning pipeline designed to analyze and detect metadata discrepancies, JPEG compression artifacts, and pixel inconsistencies in digital screenshots. The system helps verify the authenticity of transaction screenshots and digital receipts, mitigating fraud in online payments.
02 // Core Build
Key Features
- Error Level Analysis (ELA) to highlight pixel manipulation zones
- OpenCV-driven feature extraction for localized color channel analysis
- KNN classifier predicting screenshot authenticity tiers
- Structured validation pipeline evaluating JPEG compression artifacts
screenshot-forgery-detector_challenge.log
~ $cat challenge_report.txt
Accounting for varied file resolution compressions from social media apps that mask original forgery indicators.
SUCCESS:Resolved performance block using custom caching.
03 // Impact
Results
Created an offline forensic classification model with high accuracy, establishing a baseline for automatic payment receipt verification.
100%Shipped