Browser Fingerprint Lab

A cleaned public version of a bachelor-thesis prototype for collecting browser fingerprint attributes and exploring their categorical entropy.

The repository contains no collected dataset. Each installation creates its own local SQLite database under data/, and that directory is excluded from Git.

What it collects

After the user explicitly presses the collection button, the local demo records:

• User-Agent, platform and language;
• screen resolution and colour depth;
• a deterministic canvas hash;
• WebGL vendor and renderer when the browser exposes them;
• a small font-availability sample and plugin names;
• selected HTTP request headers;
• storage, cookie, Do Not Track and ad-blocking signals.

Architecture

Browser UI
   │  explicit collection
   ▼
Flask JSON API
   │
   ├── SQLite: local samples only
   └── standard-library entropy analysis

Run locally

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python app.py

Open http://127.0.0.1:8080.

The database is created automatically at data/fingerprints.s3db.

Run with Docker

docker compose up --build

The ./data directory is mounted into the container, so local samples survive a container restart but are not committed to Git.

Privacy and ethical use

This project is an educational research prototype, not a production tracking, anti-fraud or identification system. Run it only on systems you control or with informed consent. Do not publish collected datasets without an appropriate legal and ethical basis.

See docs/PRIVACY.md.

Limitations

• Browser APIs and privacy protections change over time.
• Several values may be masked, unavailable or deliberately standardised.
• The entropy numbers describe only the local collected sample.
• “Information gain” is a retained legacy pairwise heuristic and should not be interpreted as causal importance or a rigorous uniqueness score.
• The built-in development server is for local demonstration only.

Historical note

The original project was created as a bachelor-thesis prototype. This public edition removes the historical database and runtime artefacts, replaces log-file parsing with a JSON API, removes obsolete front-end dependencies, and requires an explicit user action before collection.

License

MIT — see LICENSE.