Last updated: November 2026
Quick answer: Device fingerprinting identifies a device by combining many of its browser and hardware characteristics (screen size, timezone, language, fonts, graphics rendering, operating system and more) into an identifier that stays fairly stable over time. In fraud prevention, it links accounts and sessions back to the same machine even when the attacker changes IP addresses, clears cookies or creates new emails. It is powerful against multi-accounting, account takeover and bot abuse, but it is regulated: in the UK and EU it generally needs a legal basis and transparency, and regulators have warned it is a high bar for advertising use.
Cookies are easy to clear and IP addresses are easy to change. A device’s overall configuration is much harder to fake consistently. That is why fingerprinting has become a core layer in fraud and bot detection, used by banks, marketplaces, gaming platforms and ecommerce stores.
When a page loads, a script gathers attributes that, taken individually, are common, but in combination are rare. Typical inputs include:
Server-side signals add more: the TLS and HTTP fingerprint of the connection, header order and network details. The system hashes and matches these values to produce a visitor or device ID, usually with a confidence score.
Simple open-source scripts compute a browser fingerprint that changes when the browser updates. Commercial device intelligence platforms go further: they combine fingerprinting with server-side signals, smart matching across small changes, and detection of tampering, emulators, virtual machines, incognito mode and automation tools. Vendors in this space include Fingerprint, SEON, Sardine and the device layers inside larger fraud platforms.
| Use case | How fingerprinting helps |
|---|---|
| Multi-accounting and promo abuse | Links many accounts to one device. See how to stop fake signups. |
| Account takeover | Flags logins from devices the account has never used. See how to prevent account takeover. |
| Credential stuffing | Exposes one machine testing many accounts behind rotating IPs. |
| Bot detection | Spots automation frameworks and emulators. See how websites detect bots. |
| Payment fraud | Connects a new card attempt to a device tied to earlier chargebacks. |
| Recognizing good customers | Lets trusted returning devices skip extra checks. |
For these reasons, fingerprints work best as one input in a risk score alongside network, behavior and account history, not as proof on their own.
Device fingerprinting processes information that can identify a person, so data protection law applies. In the UK and EU, storage and access rules (PECR in the UK, the ePrivacy Directive in the EU) and GDPR shape how it can be used. When Google allowed advertisers to use fingerprinting from February 2025, the UK Information Commissioner’s Office called the change irresponsible and said organizations using fingerprinting for advertising must show they meet data protection law, which it described as a high bar.
Fraud prevention is treated differently from ad tracking in many frameworks, but you still need to:
This is general information, not legal advice. Check with your privacy counsel before deploying.
It can be, when used lawfully and transparently. In the UK and EU, data protection and electronic communications rules apply, and regulators expect a clear legal basis. Fraud prevention is a common justification, while advertising use faces a much higher bar.
Partly. Privacy browsers and anti-detect tools reduce or alter fingerprints, but the measures themselves often leave detectable traces. Fraud systems treat a heavily disguised device as a risk signal.
Commercial device intelligence tools report high accuracy at recognizing returning devices, but results vary by browser, platform and how much users change their setup. Treat fingerprints as strong evidence, not certainty.
Cookies store an identifier on the device that users can delete. Fingerprinting calculates an identifier from the device’s characteristics, so clearing cookies does not reset it.