Morphed Image Detector

Machine Learning System to Detect Facial Morphing Attacks

2024 Python • TensorFlow • OpenCV • Flask
View project on GitHub

Project Overview

The Morphed Image Detector is a machine learning-based system designed to identify digitally altered facial images, particularly those used in identity fraud scenarios. This project addresses the growing concern of morphing attacks where two faces are blended to create a new identity image that can match both individuals.

Key Features

  • Advanced image processing pipeline using OpenCV
  • Custom CNN architecture trained on morphed and genuine facial images
  • Real-time detection capability through web interface
  • Detailed confidence scoring for detection results
  • Visualization of tampered regions in suspect images

Technology Stack

Backend

Python, TensorFlow, Keras, OpenCV, Flask

Frontend

HTML5, CSS3, JavaScript, Bootstrap

Deployment

Docker, Heroku (for demo)

Challenges & Solutions

Limited Training Data

Facial morphing datasets are scarce. We addressed this by creating synthetic morphed images using OpenCV and GAN-based approaches to augment our training set.

Subtle Morphing Detection

High-quality morphs are visually indistinguishable. Our solution focuses on micro-texture analysis and frequency domain features that reveal tampering artifacts.

Project Screenshots

Morphed Image Detection Interface

Web interface for image upload and analysis

Detection Results

Results showing original image as aproved

Detection Results

Results showing tampered image as denied

Detection Results

MorphShield about section

Detection Results

MorphShield creators

Detection Results

MorphShield signup page

Detection Results

MorphShield login page

← Back to Projects