Temporal-Fusion-Text

Temporal-Fusion Secure Display

🚧 Status: Experimental / Alpha

This library is currently in early development (version 0.x). It provides a working prototype of temporal-fusion display but lacks some features and optimizations.

Please consider it an alpha release. Feedback, issues, and pull requests are welcome to help improve and stabilize the package.

📄 Abstract / TL;DR

• Problem: Static screenshots (or single frames) leak sensitive text content (e.g., exam questions, proprietary documents).

• Solution: Temporal‑fusion display toggles between two glyph‑subsets at ≥2× monitor refresh rate.

  • Any single captured frame shows only partial, meaningless glyphs.
  • Human visual persistence integrates consecutive frames into coherent text.

• Use Case:

  • In online exams, this technique may help prevent cheating or information leakage via screen recording, OCR, or AI-based real-time transcription tools.

• Contributions:

  1. Pure‑JS/Canvas reference implementation with off‑screen double buffering.
  2. Automatic refresh‑rate detection & graceful degradation on low‑Hz devices.
  3. Accessibility and flicker‑safe guidelines, performance metrics, and packaging for npm.

⚠️ Threat Model

• Adversary: Any user with ability to capture screenshots, record video, or apply standard OCR tools (or real-time AI transcription).

• Assumptions:

  • Cannot physically record monitor at >2× refresh rate without specialized hardware.
  • Cannot reconstruct full content from non-integrated frames.

• Defense Scope:

  • Prevent leakage via static capture, screen-share, or simple video OCR.
  • A supplementary frontend-level method for use in lightweight DRM or exam platforms.

🧮 Algorithm & Pseudo‑code

1. Detect Refresh: estimate device refresh by two consecutive requestAnimationFrame samples.
2. Set Frequency: swapInterval = 1000 / (2 × refreshRate).
3. Partition Glyphs: text → array of glyphs; even indices → buffer A, odd → buffer B.
4. Off‑screen Buffers: draw buffers A/B once.
5. Loop: every swapInterval, toggle parity; drawImage(buffer[parity]) each rAF.

// main loop
function loop(now) {
  if (now - lastSwap >= swapInterval) {
    parity ^= 1;
    lastSwap = now;
  }
  ctx.clearRect(...);
  ctx.drawImage(buffers[parity], 0, 0);
  requestAnimationFrame(loop);
}

💻 Implementation & Usage

Installation (npm)

npm install temporal-fusion-text

API

import { TemporalFusionText } from "temporal-fusion-text";

const canvas = document.querySelector("#myCanvas");
const fusion = new TemporalFusionText(canvas, "Confidential: Q1. Solve quickly.", {
	font: "28px sans-serif",
	color: "#606060",
	opacity: 0.55,
	minSafeHz: 120,
});
fusion.run();

CLI / Demo

• demo/index.html: hosted via GitHub Pages.
• npm run start → local server.

⚖️ Limitations & Accessibility

• Low‑Hz fallback: if 2× refresh < minSafeHz, static text + warning banner.

• Accessibility:

  • Supports prefers-reduced-motion: auto‑disable fusion.
  • ARIA labels on canvas.

• Epilepsy Warning: flicker at >120 Hz could affect photosensitive users.

📦 Package & Distribution

• Module: ESM build (dist/index.js), CJS (dist/index.cjs.js).
• Types: TypeScript definitions in dist/index.d.ts.
• License: MIT.

🤝 Contributing & Roadmap

• v1.1: WebGL accelerated version.
• v1.2: Partial‑glyph OTF font generator.
• v2.0: Image‑mode (raster content).

Contributions welcome via PRs or issues on GitHub.

© 2025 Sangjune Park — MIT License