Composite fingerprint (FingerprintJS, CreepJS)
Combines dozens of signals into one visitor ID that survives cookie clearing.
In one line
Can combine many device traits into one visitor ID that survives cleared cookies
Combines dozens of signals into one visitor ID that survives cookie clearing.
Open and commercial libraries hash 50 to 100 signals, then match drifting fingerprints to past visitors. Research found 90% of fingerprints are unique on desktop but breaking up over weeks, so linking needs ML.
What it can and can't do
What the user does: Nothing. It happens when the page loads, and again on each visit
It can
- Collect about 50 to 100 small traits about your device and browser
- Merge them into one visitor ID
- Match your new profile to an old one when a few traits change
- Link your visits across days without any cookie
It cannot
- Be sure it is you, since many devices on desktop look alike or change after updates
- See your name unless you log in
- Match you well when Tor Browser or Brave hide or blur the traits
- Ask your permission first
How it differs between platforms: Fingerprints work better on desktop. Phones of the same model look more alike, so they are harder to tell apart.
Browser support
| Browser | Status | First version | Released | Note |
|---|---|---|---|---|
| Chrome | Supported | 1 | 2008-12-11 | None |
| Edge | Supported | 79 | 2020-01-15 | None |
| Firefox | Supported | 1 | 2004-11-09 | None |
| Safari | Supported | 1 | 2003-06-23 | None |
| Safari on iPhone | Supported | 1 | 2007-06-29 | None |
| Chrome on Android | Supported | 1 | Not listed | None |
| Samsung Internet | Supported | 5.0 | 2016-12-15 | None |
How a site asks for it
| API | fingerprintjs.load().then(fp=>fp.get()); CreepJS |
| How access is granted | none |
| What the user sees | Not listed |
| Needs a secure context (HTTPS) | No |
| Needs a user gesture | No |
| How long it lasts | Not listed |
| How to take it back | Not listed |
| Inside an iframe | Not listed |
Typical flow
- Collect canvas, WebGL, audio, fonts, screen, timezone, hardware
- Hash components to visitorId
- Server matches new IDs to old ones with ML
Data you get
What comes back: 128-bit visitorId; open source version about 40 to 60 percent browser-ID accuracy per vendor claims, commercial claims 99.5%
Shape: hash
| Field | Type | Unit | Range | Precision |
|---|---|---|---|---|
visitorId | hash | bits | 128 | stable weeks to months |
Technical limits
- Mobile devices are far less unique than desktop
- Shared enterprise images collide
- Browser anti-fingerprint modes create noisy or common values
The story
What it is
A script combines dozens of small browser and device traits into one ID that survives cookie clearing.
For you
You see no sign of it. Your browser quietly answers dozens of small questions, and the combined answers are often unique enough to pick you out of millions of people.
For makers
Tools like FingerprintJS and CreepJS package this as a product. Sellers use it to stop fraud, fake accounts and refund abuse. It is also used to track people across sites when cookies are blocked.
Where you see it
| Type | Example | Apps |
|---|---|---|
| security | An online bank uses a visitor ID to see that a login came from a device never seen with this customer. | FingerprintJS |
| shopping | A retailer catches one person using many accounts to abuse a first order discount. | None named |
| games | A free to play browser game stops a banned cheater from simply opening a new account. | None named |
| payments | A checkout provider scores each order partly by how familiar the device is. | Stripe Radar, Sift, Kount |
| ads | An ad network tries to link anonymous browsing across sites after the person clears cookies. | None named |
| dating | A dating site blocks scam accounts that keep coming back from the same laptop. | None named |
| analytics | A research page, Cover Your Tracks, shows how unique your browser is so you can see what leaks. | Cover Your Tracks (EFF), CreepJS |
Risks
- Hard to clear or reset
- Used without clear consent
- Can wrongly flag shared computers
Good uses and abuse
Legitimate uses
- Fraud and account-takeover detection
Known abuse
- Cross-site tracking when third-party script is on many sites
Scores
| Score | Value | Meaning |
|---|---|---|
| Sensitivity | 4 of 5 | How revealing the data is |
| Utility | 2 of 5 | How useful it is to honest apps |
| Friction | 0 of 5 | How much the user has to do |
| Fingerprint bits | 8 | Rough identifying entropy, when known |
A real example: LinkedIn
LinkedIn's script collects fingerprint signals; see research/linkedin-case.md
History
- 2018-04, paper: Hiding in the Crowd: 33.6% of 2M fingerprints unique, mobile less
- 2018-05, paper: FP-Stalker: fingerprints drift, linking gives average tracking duration up to 51.8 days
- 2023-09, Safari 17: Private Browsing noise makes composite IDs unstable
- 2025-09, Safari 26: Fingerprint protection default for known scripts
Related capabilities
Related
- Canvas 2D fingerprint. Can turn a hidden drawing into a device number that works without cookies
- WebGL render hash. Can draw a hidden 3D scene and use the pixels to tell your device apart
- Font enumeration by text metrics. Can test hundreds of font names to learn which fonts are on your device
Sources
- Gomez-Boix et al., Hiding in the Crowd (WWW 2018)
- Vastel et al., FP-Stalker (S&P 2018)
- FingerprintJS
- CreepJS
- EFF Cover Your Tracks
How well checked: Based on official documentation.