Skip to content
•

Passkeys are not a trap: my reply to The Hack

A viral Substack post urges people to drop passkeys and go back to OTP codes. It is right about governance, but it aims at the wrong target. Here is why.

cybersecurity passkey authentication phishing open-source
An open, empty rusty bear trap lit up by a glowing blue digital key with a fingerprint; in the background, in shadow, a fishing hook
An open, empty rusty bear trap lit up by a glowing blue digital key with a fingerprint; in the background, in shadow, a fishing hook

In short: The post Passkeys are a trap gets two things right: the people who write the standards do not represent users, and until 2025 passkeys really were hard to take out of an ecosystem. But it gets the mechanism wrong, because WebAuthn attestation does not reveal the operating system and synced passkeys do not provide it at all. Above all, it gets the remedy wrong: the TOTP code it recommends is exactly what today’s most widespread phishing kits steal in real time. The fear of a web that shuts out Linux users is legitimate, but it runs through other tools, and it is not fought by giving up the only widely available defence against phishing.

A post that deserves an answer

On 10 October 2026 an author writing under the pseudonym The Hack published a Substack post with an unambiguous title: Passkeys are a trap. Do not use them.

Within a few days it had collected almost 2,800 likes, over a thousand points and nearly six hundred comments on Hacker News.

The thesis goes like this: behind the convenience of passwordless sign-in hides a Trojan horse, the ability of the big platform vendors to decide which devices and operating systems are allowed to access online services. The proposed remedy is to go back to the six-digit codes generated by authenticator apps.

On the Fastal blog I wrote an analysis aimed at companies, and there I made clear it was not meant as a reply: the goal was to give clients and readers practical guidance.

This one, instead, is my reply.

I am writing it because the post strikes a chord I know well. I was on the newsgroups when Torvalds announced his “small operating system project”, and for some years now I have been giving a talk at Linux Day in Rome. The idea of a web that decides whether my operating system is “approved” worries me as much as it worries the author.

That is exactly why I think it is worth looking carefully at where the finger is pointing. Because, in my view, it is pointing at the wrong place.

Where it is right

Let me start with what holds up, because some of it does.

The governance of the standards. The author notes that no consumer organisations sit in the FIDO Alliance, the consortium that defines passkeys. That is true: the member list includes technology companies, banks, some government bodies (among them Germany’s BSI, the UK’s NCSC and NIST) and research centres. It proves no ill intent, but it is a good reason to keep one’s eyes open.

The word “passkey” is ambiguous. Very different credentials live under the same name. Synced passkeys live in the password manager (iCloud Keychain, Google Password Manager, 1Password, Bitwarden) and are replicated across all of a user’s devices: they are as secure as the cloud account that holds them. Device-bound passkeys, such as those on physical security keys like a YubiKey, never leave the hardware: they cannot be copied, and for the same reason they are lost with the device. Talking about passkeys without telling the two families apart muddles the discussion.

The lock-in was real. Until 2025 passkeys could be used from other devices, but they could not be moved in bulk from one manager to another. Those who were suspicious had good reasons.

So far, no objections. The problems start afterwards.

Where it is wrong: attestation is not what it seems

The post’s central argument is that passkeys allow sites to check the user’s operating system and firmware, and therefore to exclude Linux or any unwelcome software.

This conflates three different mechanisms.

  1. The authentication flags. On every sign-in the site receives four binary indicators: UP (the user was present), UV (the user was verified with biometrics or a PIN), BE (the credential is eligible for sync), BS (the credential is synced). They are used, for example, to require user verification. They say nothing about the operating system.
  2. WebAuthn attestation. This is an optional signed statement that identifies the make and model of the authenticator through an identifier called the AAGUID. The W3C specification describes it as evidence of “the origin of an authenticator”, and the default value of the request is no attestation at all. As Scott Helme wrote in his 14 October reply, attestation “says nothing about your OS, your browser, your extensions”.
  3. Device integrity checks. Establishing whether a phone has been tampered with is the job of mechanisms such as Google’s Play Integrity or Apple’s App Attest. They exist independently of passkeys, and many banking apps have used them for years.

Then there is a practical limit that settles the matter for the general public: synced passkeys do not provide attestation. Microsoft Entra’s documentation states it in black and white (“Synced passkeys don’t support attestation”). And Ricky Mondello, who works on authentication at Apple, reiterated it in a post published during the debate: when banks asked for attestation on syncable passkeys, platforms and browsers replied that the goal is to reduce phishing, “not some kind of trusted computing scheme”. With a closing line that leaves no room for interpretation: “over my dead body!”

The author himself, in a note added after publication, withdrew the suggestion that one of the flags signalled remote attestation of the system.

This does not mean the risk does not exist. It means it lies elsewhere. A service that wants to discriminate between devices already has the tools to do so today, and passkeys make them neither more nor less powerful. The problem is the governance of the big services, not the protocol.

Where it is most wrong: the remedy

If the analysis of attestation is inaccurate, the final recommendation is what really worries me. Going back to TOTP, the six-digit code that changes every thirty seconds, means going back to exactly the place where attackers are waiting for us.

The technique is called adversary-in-the-middle (AitM). The counterfeit site does not just collect the password: it places itself in the middle, as an intermediary in front of the genuine site. The user types password and code, the proxy relays them in real time and keeps the session cookie. Thirty seconds of validity are more than enough.

This is not a lab hypothesis. According to the Microsoft Digital Defense Report 2026, between July 2025 and June 2026 AitM accounted for 44.6% of identified phishing techniques, against 33.6% for conventional link phishing and 12.9% for malicious attachments. One of the most widespread kits, Tycoon2FA, was linked to more than 15 million emails a month in the second half of 2025, before Microsoft and Europol dismantled it in March 2026.

And then there is the telephone. In January 2026 the Google Threat Intelligence Group documented campaigns in which fake IT staff called employees to “update MFA settings”, steering them to cloned sites bearing the company’s brand. The report’s conclusion: FIDO2 keys and passkeys resist this kind of attack, push notifications and SMS do not.

The reason is simple. A passkey is bound to the domain it was created for. If I land on examp1e.com instead of example.com, the browser does not offer it: there is nothing to type, so nothing to hand over to the impostor. A TOTP code, on the other hand, I read and type back, and I can type it anywhere.

That is why NIST, in revision 4 of SP 800-63B of July 2025, states that authenticators involving the manual entry of a code “SHALL NOT be considered phishing-resistant”. CISA calls phishing-resistant MFA the “gold standard” and notes that the only widely available form is FIDO/WebAuthn authentication.

But the strongest argument comes from the author himself. In the notes added after publication, he acknowledges that passkeys are domain-bound, and therefore resistant to man-in-the-middle attacks, and that “Passkeys are definitely more secure than TOTP”.

At this point the question is unavoidable: if they are definitely more secure, why tell people not to use them?

Lock-in: an objection to update, not to shelve

On portability, the picture changed over the course of 2026.

The FIDO Alliance has defined Credential Exchange, made up of a format (CXF, published as a 1.0 Proposed Standard) and a protocol (CXP, still a draft). Since iOS and macOS 26, passkeys move between Apple’s Passwords app and managers such as 1Password, Bitwarden and Dashlane through a system picker, with no intermediate files. On Android, transfers work between Google Password Manager and compatible managers, with version requirements that vary by product.

Mondello summed it up in one line from his September article: “Nobody should feel locked into their password manager.”

That said, interoperability is not complete yet, and here open-source users have more than one reason to complain. Not every combination works in both directions. KeePassXC exports and imports passkeys in its own format, an unencrypted JSON file, but does not yet support the FIDO standard. Device-bound keys, by definition, cannot be exported.

So the lock-in criticism should be kept alive. But the way to win it is to demand that open-source managers implement Credential Exchange, not to give up on passkeys.

The right fear, the wrong target

Rereading the post, I have the impression that its real target is not passkeys. It is a future in which a few big vendors decide who is allowed to use the internet, and with which device.

It is a fear I share. But that future, if it comes, will not come from passkeys. It will come from Play Integrity and App Attest, already used today to reject phones running modified systems. It will come from the policies of individual services. It will come from market concentration.

The EFF drew the right line back in 2023: consumer-facing sites should not ask for the make and model of the authenticator. It is a sensible, well-defined battle, and it can be fought without throwing the technology away.

Telling people not to use passkeys, on the other hand, has an immediate and certain cost: leaving them exposed to the attack that works best today. The cost of the dystopian future is hypothetical; the cost of phishing is measured every day.

What I recommend, in practice

If you use online services:

  • use passkeys wherever they are available, and register at least two for important accounts, on different devices;
  • choose a manager that exports, and check that it actually does, rather than trusting the promise;
  • for critical accounts, your main email and your password manager, a physical key plus a spare is still the most robust choice;
  • turn off weak methods once you no longer need them: if password and SMS sign-in stay active next to the passkey, that is the way the attacker will come in.

If you design services:

  • do not ask consumer users for attestation: inside a company it is a legitimate control tool, on the open web it is exactly what the post fears;
  • offer passkeys as the preferred option, not the only one, and design account recovery with the same care: a weak email reset undoes everything else;
  • measure, rather than relying on other people’s numbers.

The numbers, and their scope

I will close with the data, because the debate has been full of it and it needs to be read carefully: almost all of it comes from the FIDO Alliance or from vendors, who have an interest in publicising it.

The Passkey Index 2026 reports that 93% of passkey sign-in attempts succeed, against 78% for passwords plus traditional MFA, and that 36% of successful sign-ins now use a passkey. But it is the average of 15 major consumer services, not of the whole web.

Mercari, the Japanese marketplace, records an 82.5% sign-in success rate with passkeys against 67.7% with SMS OTP, in 4.4 seconds instead of 17.

Okta, analysing more than 6,000 social engineering attempts, estimates that organisations enforcing passkeys on both sign-in and account management would have suffered no compromise at all. It is a counterfactual estimate from a vendor, and should be taken as such. The title of their reply, though, sums up mine as well: Phishing is a trap. Passkeys aren’t.

The Hack’s post had one merit: reminding us that every security technology is also a choice about who holds control. But the answer to that question is not to go back to a method we know is vulnerable. It is to adopt passkeys and, at the same time, demand portability, open standards and services that ask for no more than they need.

For the analysis aimed at companies, with the comparison between authentication methods, the scenarios for employees and customers and the European regulatory picture (NIS2, PSD2, eIDAS 2), see my article on Passkeys in the enterprise on the Fastal blog.