API Documentation

Programmatic access to transformation techniques

Base URL

https://adversarialvision.owaisshariff.com/api

Current: Low Contrast

POST https://adversarialvision.owaisshariff.com/api/transform/low-contrast

Parameters

image: string (base64) *required
text: string
contrastRatio: number (1.0-3.0)
fontSize: number
fontFamily: string
x: number
y: number

cURLExample

curl -X POST https://adversarialvision.owaisshariff.com/api/transform/low-contrast \
  -H "Content-Type: application/json" \
  -d '{"image": "data:image/png;base64,...", "text": "Hidden text"}'

PythonExample

import requests
import base64

# Read and encode image
with open("image.png", "rb") as f:
    image_data = base64.b64encode(f.read()).decode()

# API request
response = requests.post(
    "https://adversarialvision.owaisshariff.com/api/transform/low-contrast",
    json={"image": f"data:image/png;base64,{image_data}", "text": "Hidden text"}
)

All Techniques

Low Contrast

https://adversarialvision.owaisshariff.com/api/transform/low-contrast

Low Opacity

https://adversarialvision.owaisshariff.com/api/transform/low-opacity

Micro Font

https://adversarialvision.owaisshariff.com/api/transform/micro-font

SVG Path

https://adversarialvision.owaisshariff.com/api/transform/svg-path

Microtext Tiling

https://adversarialvision.owaisshariff.com/api/transform/microtext-tiling

Steganography

https://adversarialvision.owaisshariff.com/api/transform/steganography

Need Help?

For API support, integration help, or questions:

adversarialvision@owaisshariff.com

$ Adversarial Vision

// playground.tsx

// Upload Image

// Select Technique

configure(params)

1.5
More visibleLess visible
8px72px

// Canvas → low_contrast

Click and drag text to reposition

Hidden text that AI can read

// About this technique

Low-contrast text exploits the minimum perceptible difference in luminance between text and background. WCAG guidelines define 4.5:1 as minimum readable contrast for normal text. This technique uses ratios of 1.0-1.5:1, well below human perception thresholds. Human Perception: At typical viewing distance (50-70cm), the human eye requires ~3-4% luminance difference to distinguish text. At 1.2:1 contrast ratio, this drops to <2%, approaching the just-noticeable difference (JND) threshold. AI Detection: Vision models with histogram equalization, CLAHE (Contrast Limited Adaptive Histogram Equalization), or gamma correction can amplify subtle luminance differences, making sub-threshold text readable. Research: This technique demonstrates that AI preprocessing pipelines can recover information invisible to human perception. See https://github.com/NotSooShariff/adversarial-vision for detailed experiments.

→ Full research and methodology: github.com/NotSooShariff/adversarial-vision