Passwordless login is usually sold as a phishing fix, and at the point of sign-in it largely is. A passkey replaces a typed secret with a cryptographic key held on a device and unlocked by a biometric or a local PIN. There is no shared secret for a lookalike login page to harvest and no one-time code to relay in real time. For the user, the change is a fingerprint and a tap. For the security team, the phishing surface narrows sharply.

The less comfortable consequence is that the difficulty does not disappear. It moves. Once the front door is hard to pick, attackers and ordinary users alike converge on the side entrance, and in identity systems that side entrance is account recovery. A helpdesk reset, a wiped platform account or a new phone can each restore access while quietly undoing the protection the passkey was meant to provide. Recovery is becoming the real security boundary.

The failure modes concentrate in a few places

Passkeys are not one technology but a pattern. A key pair is generated on a device, the private half stays in a protected store on that device, and the public half is held by the service. Where the private key lives, and whether it can be copied, depends on the platform and its sync arrangement.

That produces a small set of recurring failure modes. The first is device loss: a phone lost, stolen or destroyed takes its passkeys with it unless they were synchronised. The second is compromise of the synchronisation account, because a passkey synced through a cloud keychain is only as strong as the account holding it. The third is migration, when a user moves between ecosystems and finds that credentials do not follow.

Each ends in the same place. The user cannot authenticate and asks for help, and the organisation has to decide whether to believe them. That decision, made many times a day in support centres, is now the weakest point in an otherwise strong authentication story.

Recovery can reintroduce the attack passkeys removed

The social-engineering attack that passkeys frustrate at login has an obvious adaptation. If the credential cannot be phished, the person holding it can be. An attacker who convinces a service desk to reset a credential does not need to defeat cryptography at all — only to supply a plausible story, a few personal details and a support agent under pressure to clear a queue.

This is not a hypothetical weakness; it is the predictable result of bolting strong authentication onto weak recovery. Legacy reset processes were designed around convenience, and a security question and a friendly conversation were considered sufficient because the credential being replaced was weak. Applying that process to a passkey lets the strongest credential in the estate be replaced through the least rigorous channel.

Identity proofing is the usual answer, and it helps, but only if it is proportionate. Verifying a government identity document and a live selfie raises the bar for an attacker, but it also raises cost, delay and privacy obligations, and it can exclude users who lack the right documents or the right camera. Recovery design is a balancing exercise between security, accessibility and support cost.

What a deliberate recovery design looks like

Organisations that take recovery seriously tend to converge on similar building blocks. The common thread is that no single method is trusted on its own, and that the path back into an account is documented before it is needed rather than improvised during an incident.

  • More than one authenticator. Registering a second device or a second passkey at enrolment gives users a route back that does not involve a human at all. Enrolment is the right moment to prompt for it, because the user is known to be present and motivated.
  • Hardware security keys for high-value accounts. For administrators, finance teams and anyone whose account can move money or change permissions, a separate physical key held offline provides a fallback that is not tied to a phone or a cloud account.
  • Recovery codes, stored deliberately. Single-use codes remain a practical last resort, provided users are told where to keep them and the codes are hashed at rest and invalidated after use.
  • Supervised identity proofing for resets. When a human has to intervene, the process should require documented evidence, a consistent script, a second reviewer for sensitive accounts and a cooling-off period before access is restored.
  • Audit trails and escalation paths. Support staff need a way to escalate a suspicious request, and the organisation needs a record of who approved what, on what evidence and when.

None of this is exotic. What distinguishes mature programmes is that these controls are treated as part of the identity architecture, with named owners, rather than as a helpdesk procedure in a knowledge-base article.

Portability is a procurement question, not just a user one

Passkey portability remains uneven. Some platforms synchronise credentials across devices within their own ecosystem; others allow export and import; some treat the credential as bound to a single device with no supported migration path. Standards work has been moving towards better transfer between providers, but implementation arrives at different times and in different forms.

For organisations, this is a procurement question. When a workforce tool adopts passkeys, the contract and the technical documentation should answer a plain question: what happens when an employee changes phone, changes platform or leaves the company? A passkey held in a personal cloud account also places recovery of a work credential under a consumer service's policies, and documenting that dependency is part of designing identity infrastructure.

Test recovery the way login is tested

Security teams routinely test authentication: they run phishing simulations, review sign-in logs, check for legacy protocols and probe for bypasses. Recovery rarely receives the same treatment, even though it is now the path of least resistance for an attacker.

A recovery test programme can be modest. Abuse-case exercises ask a support agent to handle a scripted request from someone who cannot prove who they are, and record what happened. Tabletop exercises walk through the loss of a senior administrator's device or the compromise of a synced account. Reviews check that recovery events are logged, that resets are monitored for unusual patterns and that dormant accounts with weak recovery routes are closed.

The results tend to be instructive. Teams discover that the documented process and the practised process differ, or that a fallback method quietly accepts an email link that can itself be phished.

Strong authentication at login does not raise the floor of an identity system if the route back in stays at the level it was built for.

For the Global Tech Awards, which recognises engineering across the technology sector, the topic sits where security engineering, user experience and identity management meet. Passkeys have made the sign-in step better, and that deserves credit. The harder work is to make the step after it equally deliberate, so that the protections users gain at the front door are not given away at the back.

← PreviousBack to Articles