Skip to Content
Platform.Ontology

Canvas 2D fingerprint

Draws text and shapes to a hidden canvas, then hashes the pixels the device produced.

Used to track youGraphics and fingerprintingCanvasWeb browsersWhen the page loadsWhile you use it

In one line

Can turn a hidden drawing into a device number that works without cookies

Draws text and shapes to a hidden canvas, then hashes the pixels the device produced.

GPU, driver, OS font rasterizer and anti-aliasing leave tiny pixel differences. The hash is a stable device-class identifier that works with no cookies and no prompt.

What it can and can't do

What the user does: Nothing. It happens when the page loads

It can

  • Draw hidden text and shapes, then check the pixels your device made
  • Create a number from your graphics card, fonts and software
  • Recognise the same device on later visits, even after cookies are cleared

It cannot

  • Show the drawing to you
  • Tell exactly who you are by itself
  • Get exact results from Tor Browser, which asks first, or from Brave, which adds noise
  • Be turned off by a cookie banner choice

Browser support

BrowserStatusFirst versionReleasedNote
ChromeSupported22009-05-21None
EdgeSupported122015-07-29None
FirefoxSupported22006-10-24None
SafariSupported42009-06-08None
Safari on iPhoneSupported32009-06-17None
Chrome on AndroidSupported182012-06-27None
Samsung InternetSupported1.02013-04-27None

How a site asks for it

APIcanvas.getContext('2d'); toDataURL(); getImageData(); toBlob()
Browser compat keysapi.HTMLCanvasElement.toDataURL
api.CanvasRenderingContext2D.getImageData
How access is grantednone
What the user seesNot listed
Needs a secure context (HTTPS)No
Needs a user gestureNo
How long it lastsNot listed
How to take it backNot listed
Inside an iframeNot listed

Typical flow

  1. Create off-screen canvas
  2. fillText with odd fonts plus arcs and gradients
  3. Read pixels with toDataURL or getImageData
  4. Hash the result (murmur3 or SHA-256)

Data you get

What comes back: Hash of 300x150 default canvas pixels (about 180 KB RGBA read out as a string or hash)
Shape: hash

FieldTypeUnitRangePrecision
pixelsbinaryRGBA bytes300x150 default1 level per channel before noise
hashhashbits32 to 256stable across reloads on the same stack

Technical limits

  • Pixel output is device class level, not per-device unique: millions of identical hardware and software stacks collide
  • Cannot identify the person alone; best combined with WebGL, audio and fonts
  • Tainted cross-origin canvases refuse readback

The story

What it is

A script draws text and shapes in a hidden canvas, then hashes the pixels the device produces.

For you

You do not see it. The same drawing looks a little different on different computers because of fonts and graphics chips, so your result works like a quiet ID card.

For makers

Ad tech and fraud vendors use it as one core signal in a bundle. Its value is that it works on most browsers with no permission and stays the same after cookies are cleared.

Where you see it

TypeExampleApps
adsA data broker uses the canvas hash to match your phone visit and laptop visit to the same profile.None named
securityA payments site flags a checkout when the same canvas hash shows up on hundreds of fake orders.None named
gamesA game site bans a cheater and then recognizes the same device when they try to make a new account.None named
analyticsA web analytics tool without cookies uses the canvas hash to count unique visits.None named
socialA social site blocks a banned user from signing up again from the same device.None named
shoppingA ticket seller finds a single scalper using one browser behind many accounts.Ticketmaster
bankingA bank links your new login to your trusted device list without asking for a code every time.None named

Risks

  • Cookie clearing does not help
  • Works across sites

What the platform says

Good uses and abuse

Legitimate uses

  • Anti-fraud device ID, bot detection, font and emoji support checks

Known abuse

  • Cookieless cross-site tracking, fingerprint resale, covert re-identification after cookie clearing

Scores

ScoreValueMeaning
Sensitivity4 of 5How revealing the data is
Utility1 of 5How useful it is to honest apps
Friction0 of 5How much the user has to do
Fingerprint bits7Rough identifying entropy, when known

History

  • 2012-05, paper: Mowery and Shacham 'Pixel Perfect' introduces canvas fingerprinting
  • 2014-07, paper: Acar et al. find AddThis canvas fingerprinting on 5% of top 100k sites
  • 2015-09, Firefox 41: Canvas extraction prompt added under privacy.resistFingerprinting (Tor Uplift)
  • 2023-09, Safari 17: Canvas noise in Private Browsing (Advanced Fingerprinting Protection)
  • 2025-09, Safari 26: Fingerprinting protection extended to all browsing for known fingerprinting scripts
  • 2025-11, Firefox 145: Randomized canvas readback in Private Browsing and ETP Strict

Sources

How well checked: Based on official documentation.

On this page