Breaking Lab srl
The Passkey Promise Under Pressure
Passkeys have been relentlessly promoted as the future of authentication — a phishing-resistant, password-free standard that was supposed to close the door on credential theft once and for all. Major platforms including Google, Apple, and Microsoft have spent the last two years encouraging users to migrate. But new research from Palo Alto Networks’ Unit 42 threat intelligence team reveals a set of attack techniques that can undermine passkey-protected accounts without requiring a single fingerprint scan, PIN entry, or visible prompt on the victim’s screen.
Three Attack Paths, One Common Thread
Unit 42 identified three distinct exploitation routes — which the researchers have named Pass-ta-key, Silver Pass-ta-key, and Golden Pass-ta-key — each targeting Chrome’s Google Password Manager cloud authenticator, the component responsible for storing and syncing passkeys across a user’s devices. The attacks do not require administrator privileges; malware running at ordinary user-level permissions on a Windows machine is sufficient to initiate them.
The most severe variant, Golden Pass-ta-key, targets the master key that Google Password Manager uses to protect locally stored credential material. By extracting or abusing this key, malware can effectively impersonate the legitimate user during an authentication flow, bypassing the biometric or PIN verification that passkeys are designed to enforce. Crucially, the victim sees nothing — no dialog box, no authentication request, no system alert. The sign-in happens silently in the background.
Why This Is More Than a Browser Bug
It is tempting to frame these findings as simply another browser-level vulnerability, but the implications run deeper. Passkeys rely on a chain of trust that includes both the cryptographic protocol (typically FIDO2/WebAuthn) and the authenticator platform — in this case, Google’s own credential manager. The protocol itself is not broken; what Unit 42 is demonstrating is that the local software layer responsible for gatekeeping access to passkey material can be compromised by malware that has already gained a foothold on the device.
This distinction matters enormously for enterprise security teams. An attacker who deploys an infostealer or a remote access trojan — increasingly common thanks to phishing campaigns and supply-chain compromises — could pivot directly to passkey-protected corporate accounts without the additional step of convincing the user to hand over a one-time code or password. The “something you have” and “something you are” factors that passkeys are supposed to represent both collapse once the device itself is compromised.
Google’s Response and Current Mitigations
At the time of writing, Google has not issued a public patch specifically addressing the three techniques described by Unit 42, though the company was notified as part of responsible disclosure procedures. Google has previously argued that device-level compromise falls outside the security boundary that passkeys are designed to protect — a position that is technically defensible but may be cold comfort to organizations deploying passkeys as a primary MFA replacement.
In the interim, security practitioners are recommending a layered approach: ensuring endpoint detection and response (EDR) tooling is deployed and current, auditing which applications have access to credential storage on Windows endpoints, and avoiding sole reliance on any single authentication mechanism regardless of how modern it appears.
A Reminder That Security Is Systemic
The Unit 42 research is a timely corrective to the narrative that passkeys represent an unconditional leap forward in account security. They are unquestionably superior to passwords in resisting phishing and remote credential theft — but no authentication technology operates in a vacuum. When the device executing the authentication is compromised, the security guarantees weaken dramatically. As the industry accelerates passkey adoption across consumer and enterprise platforms, the findings underscore that endpoint hygiene, malware detection, and zero-trust architecture remain as critical as ever. Replacing one weak link in a chain is progress; mistaking it for an unbreakable chain is a different matter entirely.
