Browser Fingerprint Test
19 cookieless signals any site can read, scored for how identifying they are, plus the bot and tampering checks a fraud system runs on them.
Bot and tampering checks
A fingerprint says which browser this is. Fraud and bot-defence scripts also ask whether a person is driving it and whether its answers were edited. These are the checks they run, read in this tab.
Running checks
Automation flags, values that must agree, patched built-ins and VM hints.
What a browser check sees, and what needs a server
This page runs the client half in the open. Device intelligence and bot-defence services pair it with a server half that a page cannot run alone.
| Layer | Readable in this tab | Needs a server | Why fraud teams use it |
|---|---|---|---|
| Device signals | Yes: canvas, audio, GPU, fonts, screen, locale | Matching them to a visitor seen before | Links visits with no cookie, after a cookie clear or in a private window |
| Automation traces | Partly: stealth plugins patch the obvious flags | Re-checking from request data the page cannot edit | Bots drive sign-up, checkout, scraping and credential stuffing |
| Tampering and spoofing | Yes, the checks above | Signing the result so a script cannot replay a clean one | Anti-detect browsers sell a fresh, clean profile per account |
| IP reputation, VPN, proxy, Tor | Your exit address (Exit IP lookup) | Datacenter, residential proxy and relay lists | Proxies hide many accounts behind fresh addresses |
| TLS and HTTP fingerprint | No | Yes: JA4 hash, header order, HTTP/2 settings | Catches scripts that copy a browser user agent but not its network stack |
| Identity across weeks | One hash kept in this browser | Fuzzy matching that survives browser updates | A stable ID through updates is the hard part, and where accuracy is won or lost |
| Velocity across accounts | No | Counts per device, per IP, per card | Many accounts on one device is the core multi-accounting and promo-abuse signal |
How the browser fingerprint test works
This browser fingerprint test reads up to 19 signals any website can collect from your browser without a cookie, scores how identifying they are together, and checks your clock and language against your VPN exit. The leak test and exit IP lookup run from the same exit snapshot in the panels beside it.
The page runs the same reads a tracking script would, in this tab, and lists the answers. It draws a hidden canvas and renders a short audio clip offline, each twice, and hashes the result. It reads the GPU name and WebGL limits, checks which of 72 common fonts are installed by measuring text width, and reads the screen size, display and accessibility preferences, timezone, languages, regional formats, CPU threads, memory, touch points, platform, high-entropy client hints, user agent, connection estimate, installed speech voices, the count of cameras and microphones, Do Not Track and Global Privacy Control, storage settings and a few floating-point results from the JavaScript engine.
The bot and tampering panel reads navigator.webdriver, headless names, driver variables, plugins, permission answers and window size; compares the user agent with client hints, navigator.platform, the JavaScript engine and touch support; checks the clock offset against the timezone name; tests whether 15 built-in getters and functions still report native code; reads the GPU string for software renderers and hypervisors; and, while you stay on the page, scores mouse path straightness, the timing between moves and keys, and any input events a script dispatched. The "Last visit" column compares each signal with the copy this browser kept from your previous run.
It then takes your exit address, the one value a tunnel changes, from the run's shared snapshot of the pwnsy VPN API, and compares your clock and language regions with the exit's timezone and country.
How to read the result
- Points per signal: strong 3, moderate 2, weak 1. A blocked signal scores 2, since a missing API is rarer than a common answer. A signal randomised per read (two reads differ) scores 0, because it cannot link two visits.
- Identifiability score is the sum, shown against 3 points for every signal read. 70% or more reads red, 45% or more amber, below that green.
- Strong, moderate and weak are our judgement of how identifying each signal usually is. They are not measured against a population of visitors: that would mean storing everyone's fingerprint, which this page does not do.
- Combined hash covers every signal that read cleanly (state ok), leaving out the connection estimate and any signal that was blocked or randomised. Same as last visit compares it with the copy kept in this browser's localStorage only, which shows what a site can do with no cookie at all.
- Agrees or mismatch under the tiles compares your clock and language regions with the exit. A mismatch is something a site can see while you use a VPN.
A VPN changes only the exit address. Tor Browser makes every user look alike, which works best. Firefox resistFingerprinting does some of the same, and Brave randomises canvas and audio per site and session. Blocking one signal often makes you rarer.
Browser fingerprint FAQ
Can a website tell if my browser is automated?
Often. Selenium, Playwright and Puppeteer leave traces: navigator.webdriver set to true, a HeadlessChrome name, driver variables on the page, no PDF plugins or a window with no frame. Stealth plugins hide the obvious ones, so sites also check whether values agree with each other and whether built-in functions were replaced.
Why would a site flag a mismatch between my user agent and client hints?
Spoofing tools often edit the user agent string and leave the client hints, navigator.platform or the JavaScript engine unchanged. When they disagree, the browser is claiming to be something it is not, which fraud systems weigh heavily.
Which signals survive clearing cookies?
All of them. Canvas, audio, GPU, fonts, screen and locale come from the device, so clearing cookies or opening a private window leaves them. They change with a browser update, a new driver or new hardware, which the Last visit column shows signal by signal.
Does a VPN hide my browser fingerprint?
No. A VPN changes only the exit address. Canvas, audio, GPU, fonts, screen, timezone and language are read from your device, so they stay the same with the tunnel on or off.
Is my fingerprint sent to pwnsy?
No. Every signal is read in this tab and none is sent. The page reads your exit address once from the pwnsy VPN API for all three panels, and the leak test panel sends its own DNS probes; neither carries anything measured here. The combined hash is kept in this browser's localStorage so your next visit can compare against it.
Why does a randomised signal score zero?
The page reads the canvas and audio hashes twice. When the two reads differ, the browser is adding noise, so that value cannot link two visits and scores 0 points.
Why does a blocked signal still score 2 points?
A missing API is rarer than a common answer, so blocking one signal can make a browser stand out more.
What actually reduces a browser fingerprint?
Tor Browser makes every user look alike, which works best. Firefox resistFingerprinting does some of the same, and Brave randomises canvas and audio per site and session.
Every value is read by this page in your browser. Only the exit address comes from the pwnsy VPN API, and nothing measured here is sent to it.