Image Perturber

Defeats both cryptographic (SHA-256) and perceptual hash (dHash) duplicate detection systems. Should pass most image duplicate checking software tests.

Real-time similarity analysis shows effectiveness

🖼️

Drop your image here

or

Supports JPG, PNG, WebP, and other image formats

How it works

  • Iterative Optimization: Runs 50 iterations applying smooth perturbations
  • L2 Regularization: Keeps changes minimal and imperceptible
  • Gaussian Smoothing: Ensures zero visible artifacts or patterns
  • Early Exit: Stops automatically at <65% similarity
  • Multi-Scale Noise: Smooth noise at different frequencies
  • Dual Hash Testing: Tests both SHA-256 and dHash algorithms

🎯 Mimics Dolhansky's & Ferrer's approach: Iterative gradient descent with L2 regularization produces imperceptible smooth perturbations that defeat perceptual hashing.