Passkeys have spent the last two years being marketed as the technology that finally kills phishing: no password to steal, no code to intercept, cryptographic proof tied to the device itself. Unit 42, the threat research arm of Palo Alto Networks, has now published three attacks against Chrome’s Google Password Manager on Windows that quietly undercut that promise, not by breaking the cryptography, but by walking straight through the software that manages it.

Three ways in, one theme

The researchers call the family Pass-ta-key, and it comes in three escalating forms. The basic version has malware running with ordinary user privileges extract the TPM-backed identity key that Chrome uses to authenticate to Google’s Cloud Authenticator, then use standard Windows Cryptography APIs to sign a login assertion. Google accepts it as valid, no fingerprint, PIN or user-verification prompt required, because the request looks exactly like a legitimate one from the same machine. Silver Pass-ta-key extends this to a broader set of protected operations, and Golden Pass-ta-key, the most damaging of the three, retrieves the security domain secret, the master key Google temporarily sends to Chrome during device registration or account recovery that encrypts every passkey synced through that account. Steal that once and every passkey the victim has ever created, across every site that supports them, is exposed.

Why “requires a TPM” is not the reassurance it sounds like

The attacks specifically target Windows machines with a Trusted Platform Module, the hardware component meant to keep exactly this kind of key material out of software’s reach. That is precisely what makes the disclosure notable: TPM-backed protection is the baseline security story every passkey vendor tells enterprise buyers, and Unit 42 demonstrated that user-level malware, not a kernel exploit or physical access, was enough to get past it. The researchers reported the findings to Google and separately flagged related user-verification weaknesses to services including eBay before publishing.

What this means for organisations betting on passkeys

Passkeys still beat passwords against remote phishing and credential-stuffing, and nothing here changes that. What it does change is the assumption that a passkey removes endpoint compromise from the threat model. If a European organisation adopted passkeys to satisfy NIS2 authentication requirements or PSD2 strong customer authentication, and treats that adoption as complete, this research is the argument for revisiting it: endpoint detection and response on the devices holding those keys is not optional infrastructure sitting next to the passkey rollout, it is the control that determines whether the rollout actually holds under attack. Enterprises should also ask their identity provider what user-verification enforcement looks like in practice, since the gap Unit 42 found is precisely that verification prompts can be bypassed rather than merely guessed.

If your organisation is rolling out passkeys, evaluating passwordless authentication for NIS2 or PSD2 compliance, or wants a review of how your endpoint security posture holds up against credential and session theft, contact Excello Digital. We help European businesses design authentication architectures that survive contact with real malware, not just marketing claims.