Passkey enrollment fraud signals matter because passkeys are usually treated as a stronger, lower-friction login method. That is good for legitimate users, but it also creates a policy question: what if the session adding the passkey is not trustworthy?
A passkey is a public-key credential bound to an account after some existing authentication step. If a hijacker gets into the account first, enrolling a new passkey can turn temporary access into persistent access. Your job is not to block every new device. Your job is to decide when enrollment should be allowed, challenged, delayed, or reviewed.
Start with the enrollment moment
Treat passkey enrollment as its own risk decision, not just a setting inside the account page. The user may already be logged in, but that does not mean the current browser agent, device, and session context are normal for that account.
At minimum, evaluate the request server-side before creating the credential. Send your device identification event with a linkedId such as the account ID and keep the requestId attached to the passkey enrollment attempt. That gives your fraud and support teams a clean trail: which account, which device, which session, and which decision.
A useful policy starts with one question: is this device expected for this account, or is it new in a way that changes the risk?
Check device and browser signals first

New or unstable visitorId
A new visitorId is not automatically bad. Users buy phones, reinstall browsers, clear storage, and use work devices. The risk increases when the account has only used stable devices before and the first new device immediately tries to add a passkey.
Look for instability too. If the same session produces changing identifiers, missing browser attributes, or signs of tampering, enrollment deserves a stronger response than a normal login.
Device reuse across accounts
accountsOnDevice helps answer a different question: has this device touched other accounts? That matters for account abuse, recovery abuse, and support-driven takeover patterns.
For passkey enrollment, a high account count should usually move the decision from allow to challenge or review, especially when the device has not been linked to this account before. Keep the policy narrow: device reuse alone may reflect a family computer, QA team, or shared office machine.
Smart Signals that change the response
Smart Signals such as bot behavior, VPN, Tor, incognito, tampering, or VM should not replace your policy. They should adjust it.
For example, a known device using a VPN may still be safe enough after reauthentication. A new device with automation, tampering, and a fresh password reset is a very different enrollment event. Evaluate the overall decision first, then add reason codes that explain why the enrollment was challenged or blocked.
Compare the session to the account history
Device signals are strongest when you compare them to session history. A passkey enrollment attempt is higher risk when it follows a sensitive account event: password reset, email change, recovery code use, MFA reset, support-assisted recovery, or a login from an unfamiliar browser agent.
Build a short lookback window for sensitive events. You do not need to keep everything forever. You need enough context to answer whether this enrollment is happening during a fragile account state.
Useful checks include:
- Was the session recently created after a recovery flow?
- Did the user change email, phone, MFA, or password in the same session?
- Is the enrollment coming from a new
visitorIdfor this account? - Does the device have many linked accounts or recent failed attempts?
- Are bot, Tor, VM, incognito, or tampering signals present?
The result should be a decision with reason codes, not a vague risk label.
Gate enrollment without blocking good users
A good passkey policy has more than allow and block. Most legitimate new-device cases should still have a path forward.

Use a tiered response:
- Allow when the device is known, the session is stable, and there are no meaningful risk signals.
- Step-up challenge when the device is new but the account context is otherwise normal. Ask for current password, existing MFA, email confirmation, or a recent trusted-device approval.
- Delay when the session recently used recovery or changed key account fields. You can let the user continue using the account while postponing passkey enrollment.
- Review or block when multiple signals stack together, such as new device, tampering, bot behavior, recent recovery, and device reuse across many accounts.
Avoid policies that punish every new device. The goal is to protect the binding of a durable credential, not to make account security unusable.
Implement the decision server-side
Run the decision on your server before storing the public key credential. Do not rely on client-side flags alone. The browser can collect context, but your server should evaluate the requestId, account state, and enrollment policy.
A simple policy can look like this:
risk = []
if device.isNewForAccount(accountId):
risk.add('new_device')
if device.accountsOnDevice > policy.sharedDeviceLimit:
risk.add('device_reuse')
if smartSignals.bot or smartSignals.tampering or smartSignals.vm:
risk.add('host_integrity')
if session.afterRecovery or account.recentSensitiveChange:
risk.add('fragile_session')
if risk == []:
allow_enrollment()
elif risk.contains('fragile_session') and risk.contains('new_device'):
delay_or_step_up()
elif risk.length >= 3:
block_or_review()
else:
step_up_challenge()
Store the decision, reason codes, visitorId, and requestId with the enrollment attempt. If a user later reports account abuse, support can see why the credential was allowed or challenged without guessing.
FAQ

Should every new device be challenged before passkey enrollment?
Not always. A new device is common and often legitimate. Challenge when the new device appears with other risk signals, such as recent recovery, automation, tampering, or unusual account reuse.
Is incognito mode enough to block enrollment?
No. Incognito is a signal, not a verdict. It may be normal for privacy-conscious users. Use it to raise friction when combined with new-device or session-risk signals.
How long should enrollment signals be retained?
Keep the minimum data needed for your fraud, security, and support policy. Document retention, support erasure requests where required, and have counsel review privacy-sensitive decisions.
Try it free
Prynt helps you identify devices, detect bots, and make server-side fraud decisions with privacy-friendly controls and deployment options you can run under your control. If passkey enrollment needs better device context, try Prynt.
Try it free
Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.