humanoid.js

Zero-dependency, single-file web tool that scores how human-looking a pointer/touch interaction is — and tells you which specific signals gave it away.

👉 Live demo (or open index.html in any browser)

---

What it does

Every click, long-press, or swipe on the test pad is analyzed in real time across ~30 signals that click-fraud and anti-bot systems actually look at:

Category	Signals
Event trust	isTrusted, pointer/touch source, navigator.webdriver
Pressure	min / max / avg / stddev, change count, Force Touch
Contact geometry	radiusX, radiusY, rotationAngle, width × height
Trajectory	straightness, path efficiency, curvature, jitter, Bezier-match
Time profile	duration, sample interval avg/stddev, pauses
Advanced	entropy, angle stddev, speed consistency, coordinate decimal precision
Environment	WebDriver, HeadlessChrome, Selenium markers, UA / touch mismatch, etc.

The output is a 0–100 risk score (higher = more human-like) plus a list of exactly which rules fired, shown directly on the page.

Usage

No build step. No dependencies.

# just open it
open index.html            # macOS
xdg-open index.html        # Linux

# or serve locally
python3 -m http.server 8000

Page defaults to English; click the EN / 中文 toggle in the top-right to switch (preference is remembered in localStorage).

Repository layout

.
├── index.html              # the whole app — HTML + CSS + JS, no build
├── humanoid-example.png    # README screenshot
├── LICENSE
└── README.md

Contributing

Issues and PRs welcome. Please keep:

• Single-file (index.html) with no build tooling.
• No runtime dependencies.

License

MIT — see LICENSE.

Disclaimer

For research, testing, and defensive/awareness purposes. Do not use to bypass anti-fraud or anti-abuse systems you are not authorized to test.