Tue 11 Aug 2026 / 12:07 ET
Kernel
Internet 3 min read

Pass-ta-key attack puts Google passkeys’ Windows threat model under scrutiny

Unit 42 says malware on infected Windows PCs can abuse Google’s synced passkey flows, while critics say the core risk is endpoint compromise.

June Castellano

By June Castellano / Platforms & Power Reporter

Pass-ta-key attack puts Google passkeys’ Windows threat model under scrutiny
img: Ars Technica

The Pass-ta-key attack is a set of techniques that Palo Alto Networks’ Unit 42 says can let malware already running on a Windows computer abuse Google Password Manager’s synced-passkey system. The research, published August 3, describes ways to impersonate a trusted device, interfere with user-verification setup, or recover material protecting synced passkeys. The catch is not decorative: every reported technique starts with a compromised Windows endpoint.

That requirement defines the story. Unit 42 calls the work a new attack surface in passwordless authentication. Ars Technica argues that malware controlling a logged-in Windows machine has long been able to raid other authenticated applications, making the underlying risk broader than passkeys.

Neither account settles the engineering questions by declaring victory. The research points to workflows around Google’s cloud-synced passkeys and desktop Cloud Authenticator, rather than showing that a stranger can remotely steal passkeys from a clean device.

What does the Pass-ta-key attack do?

Unit 42 says the first technique, which it calls Pass-ta-key, allows unprivileged malware to pose as a trusted device and obtain valid passkey authentication without administrator rights, user interaction, or device unlocking. A separate technique reportedly registers an attacker-controlled user-verification key with Google’s Cloud Authenticator. The third reportedly obtains a key used to protect synced passkeys after related secret material was exposed through Chrome’s internal FIDO logs or remained in process memory.

Those are researchers’ claims, not a published response from Google. The evidence available does not include an affected-version list, a CVE, patch information, independent reproduction, or Google’s account of the findings. Unit 42 said the attacks could misuse onboarding, recovery and device-trust processes to take over passkey-protected accounts, bypass user verification and extract synced private keys.

Why Windows and cloud syncing are part of the story

A passkey uses public-key cryptography: a service receives a signed authentication assertion, while the private key is meant to remain under the user’s control. The design removes the shared password that can be reused, phished or copied from a breached service, according to Unit 42 and Ars Technica.

FIDO2 standards do not require passkeys to live in a TPM or another dedicated hardware enclave, Ars Technica reported. For Google Password Manager on Windows, passkeys are held as end-to-end-encrypted cloud data, the outlet said. In its description of the login flow, a user or device key from the TPM, plus an already authenticated Google account, causes Google’s backend authenticator to produce the assertion signed with a private key held in encrypted form on Google’s server.

That cloud design helps users sync credentials between devices, but it also makes device onboarding and trust decisions security-critical. Ars Technica reported that Google Password Manager, 1Password and Dashlane use encrypted cloud-stored passkey blobs on Windows, while their implementations on other platforms store passkeys locally.

Does this mean passkeys are broken?

No evidence here shows conventional phishing, password reuse, credential stuffing, or a service-side password breach defeating passkeys. Those are the threats passkeys are designed to reduce because the user does not hand a reusable secret to a site.

The reported weakness is the uglier, familiar boundary: an infected computer that is already authenticated to sensitive accounts. Ars Technica’s analysis says that such malware can generally operate within the user’s access and abuse credential-management functions. Unit 42’s work identifies specific Google synced-passkey workflows that researchers say make that control more consequential.

The practical conclusion is narrower than the attack name suggests. Passkeys still remove password-era failure modes. They do not turn a malware-controlled, signed-in Windows machine into a trustworthy authenticator. That distinction is the part worth keeping after the pasta pun has gone stale.

This story draws on original reporting from Ars Technica.

More Internet/

view all ↗