How a browser fingerprint differs from cookies
| Cookies and other storage | Browser fingerprint | |
|---|---|---|
| Where it is stored | In your browser | Not stored; the website reads and computes it on every visit |
| Can you clear it? | Yes, by clearing it or using incognito mode | No; on the same device it usually stays stable for a long time |
| Common uses | Keeping you signed in, shopping carts, preferences | Device recognition, fraud prevention, risk control, ad attribution |
Common fingerprint signals
| Signal | What it reveals | Common inconsistencies |
|---|---|---|
| User-Agent and Client Hints | Browser brand and version, operating system, architecture | The UA says Chrome 144, but Client Hints or a Worker report a different version |
| Screen | Resolution, available height, color depth, device pixel ratio | Available height does not match the system taskbar |
| Language and time zone | navigator.language, Accept-Language, Intl time zone | Time zone does not match the IP location |
| Fonts | List of installed fonts and their rendered widths | A Windows UA, yet macOS-only fonts are detected |
| Canvas | Hash of rendered text and graphics | Changes on every read (random noise) or does not match the GPU |
| WebGL / WebGPU | GPU vendor and renderer, parameters, pixel output | A Windows system reports an Apple GPU |
| Audio | Hash of the AudioContext processing output | Results vary between runs on the same device |
| ClientRects | Subtle differences in element dimensions | The main window and iframes return different results |
| Hardware | CPU core count, device memory | Core count clearly does not match the device model |
| WebRTC | Local and public IP candidates | The real public IP is exposed even when a proxy is used |
| Automation traces | navigator.webdriver, headless browser markers | Properties left behind by automation tools |
Why “consistent” matters more than “random”
On a real device, the characteristics agree with each other and stay stable over time. That is why test sites look not only at whether a single value is rare, but at whether all the values can be true at the same time:
- Consistent across signals: the OS, fonts, GPU, screen and browser version must come from the same kind of device.
- Consistent across contexts: the main page, iframes and Workers should see the same identity.
- Stable over time: the same device produces the same fingerprint on every visit; “random noise” that changes every time is itself a red flag.
- Consistent with the network: time zone, language and geolocation should match the location of the exit IP.
How to test your own browser fingerprint
| Tool | Focus |
|---|---|
| CreepJS | Consistency across contexts, Canvas, WebGL, Audio, fonts, Intl and more, with suspicious “lies” flagged |
| Pixelscan | Consistency of browser, OS, location, automation and fingerprint masking |
| BrowserLeaks | Raw data for Canvas, WebGL, fonts, WebRTC, Client Hints and more, item by item |
| IPhey | An overall assessment of browser, location, IP, hardware and software |
Test two profiles side by side
An antidetect browser is useful because profiles differ from one another; testing a single profile will not reveal problems.
Test again after a restart
Close the profile and open it again; key values (Canvas, WebGL, time zone) should stay the same.
Compare with a real browser on the same computer
Some test pages flag even the official Chrome on the same computer in red, so first work out which warnings simply reflect how the test site judges browsers.
Record fields, not colors
Note the test time, OS, browser and kernel versions, the proxy and the exact fields that failed. Detection rules change, so a result only reflects that moment.
How Prism handles these signals
Prism applies fingerprint settings at a low level in its custom-built Chromium kernel instead of injecting scripts into pages. The OS, CPU, screen and GPU come from the same hardware template, and details such as Canvas, WebGL and Audio are fixed by each profile’s fingerprint seed. For item-by-item details, see Fingerprints and hardware templates.
FAQ
- Does clearing cookies or using incognito mode change my fingerprint?
- No. The fingerprint comes from characteristics of the device and the browser itself; clearing storage does not change it.
- Does a proxy change my browser fingerprint?
- A proxy only changes the network exit. Time zone, language, WebRTC and other signals can still reveal mismatches, so configure them together with the fingerprint.
- Does a more unique fingerprint make me easier to identify?
- Rarer combinations are easier to single out, but self-contradictory combinations matter more. A common, complete hardware combination is safer than randomly pieced-together values.
- If a test page shows red, is something definitely wrong?
- Not necessarily. Look at the specific fields first and compare with a real browser on the same computer. A detected proxy, font differences or the site’s own rules can all trigger warnings.