Pass-ta-key Attack: What It Really Reveals About Passkey Storage
Original: New Pass-ta-key attack reveals all the things we didn't know about passkeys
Why This Matters
Highlights a critical but widely misunderstood gap in passkey security assumptions on Windows platforms.
Researcher Arie Olshtein of Palo Alto Networks demonstrated the 'Pass-ta-key' attack, which extracts all passkeys stored in Google Password Manager for Windows on malware-infected machines. The attack is not novel but exposed a widespread misunderstanding: most platforms do not store passkeys in TPMs.
Security researcher Arie Olshtein of Palo Alto Networks published findings last week on a technique dubbed 'Pass-ta-key,' showing that passkeys stored in Google Password Manager (GPM) on Windows can be fully extracted when the machine is infected with malware. The disclosure sparked confusion among users who believed passkeys are always stored in the Trusted Platform Module (TPM), a hardened silicon enclave for cryptographic keys.
In reality, FIDO2 specifications maintained by the FIDO Alliance do not require passkeys to be stored in TPMs or any dedicated secure hardware. Nearly all platforms—including macOS, iOS, and Android—store passkeys locally on the device in software. Microsoft is the notable exception, offering users the option to store passkeys in the Windows TPM, though it primarily recommends this for enterprises.
The shift to local software storage was a deliberate design decision made to enable cross-device sync. Storing passkeys in TPMs made syncing impractical, as keys would need to be individually recreated on each new device. FIDO architects concluded local storage was acceptably safe because OS-level app permission sandboxing prevents malware from accessing private keys—on macOS, iOS, and Android, this isolation holds unless a zero-day exploit compromises the OS itself. The Pass-ta-key attack targets Windows specifically, where GPM's storage model differs.