EchoPrint AI

M1–M5 fingerprint lab — separates empty Chrome from hardened browsers.

Live: echo-print-ai.vercel.app · Docs: ARCHITECTURE-M1-M5

---

Modules

#	Name	What it does
M1	Network (/api/fp)	IP ASN/type/VPN, JA3/JA4 hooks, header order, WebRTC vs HTTP, geo↔timezone map (>1000 km = mismatch)
M2	Hardware	stable_id: 3× canvas, WebGL, WebGPU, Audio, fonts, screen, Math — same on all browsers of one PC
M3	Software	Spoof (UA/CH/GPU/fonts), adblock DOM, Brave, canvas noise, tracker script probes → protection 0 vs 95
M4	Four scores	A uniqueness · B spoof · C aggressiveness · D vulnerability + trackability %
M5	Advanced	localStorage temporal ID, emoji FP, VM WebGL strings

History & compare — last 10 full reports stay in localStorage only. Select two scans (e.g. stock Chrome vs Brave) and diff scores / stable_id / protection.

Privacy: M2–M5 run client-side; M1 is an ephemeral server request (no DB). History never leaves the browser. JA3 needs edge headers or sidecar.

---

Quick start

git clone https://github.com/Evreu1pro/EchoPrint-AI.git
cd EchoPrint-AI
npm install
npm run dev

Open http://localhost:3000.

npm run build
npm start
npm test          # unit tests (scoring, compare, IP extract)
npm run typecheck
npm run lint

---

Deploy

Vercel (recommended)

1. Import Evreu1pro/EchoPrint-AI in Vercel
2. Framework: Next.js (auto)
3. Build: npm run build · Output: default

Static export (GitHub Pages / Netlify)

In next.config.ts:

const nextConfig = {
  output: 'export',
  images: { unoptimized: true },
  // basePath: '/EchoPrint-AI', // if project site on GitHub Pages
};

Then npm run build → publish the out/ folder.
Note: /api/fp will not work on pure static hosting (M1 falls back to empty network module).

---

Architecture

src/
  app/                      # Next.js App Router UI + /api/fp
  components/scanner/       # Report, history, compare
  hooks/useModuleScan.ts    # Pipeline + local history
  lib/
    modules/                # M1–M5 pipeline
      m1-network/           # IP extract, intel, geo↔tz
      m2-hardware/
      m3-software/
      m4-scoring/
      m5-advanced/
    history/                # localStorage + compare (pure)
    i18n/messages.ts        # EN / RU

See docs/MODULE-1-NETWORK-DETECTIVE.md and docs/ARCHITECTURE-M1-M5.md.

Adding a detector

1. Collect in the matching src/lib/modules/mN-*/ module
2. Extend types.ts if the report shape changes
3. Wire scoring in m4-scoring/scores.ts when the signal should move A/B/C/D

---

Methodology (short)

Uniqueness — estimated signal entropy → score 0–100 (higher = easier to re-identify).

Spoof / undercover — geo↔tz mismatch, datacenter/TOR, WebRTC≠HTTP, software spoof findings.

Protection / aggressiveness — Brave, canvas noise, adblock DOM, tracker script probes.

Vulnerability — open ad surface + powerful browser APIs (USB, Bluetooth, GPU, …).

This project is educational. It does not claim perfect bot detection or legal tracker blocking.

---

Stack

• Next.js 16 · React 19 · TypeScript
• Tailwind CSS 4 · shadcn-style Button
• Vitest (unit) · GitHub Actions CI

---

License

MIT

---

Built for privacy education. Contributions welcome.