Synced passkey attacks: what RPs still control
Why passkeys changed the attack surface At Authenticate 2026, Palo Alto Networks showed a live phishing campaign against iCloud Keychain recovery and three endpoint attacks against Googleβs synced passkey design. None
Why passkeys changed the attack surface
At Authenticate 2026, Palo Alto Networks showed a live phishing campaign against iCloud Keychain recovery and three endpoint attacks against Googleβs synced passkey design. None of them breaks WebAuthn or FIDO2 cryptography. That matters, because a lot of teams still ask the wrong question: βCan passkeys be phished?β
At the protocol level, classic phishing is largely gone. The real problem is everything wrapped around the passkey: account recovery, cloud sync accounts, endpoints, and fallback authentication that attackers can still use.
That is the core shift in the passkey threat model. Synced passkeys trade hardware binding for recoverability. They are much easier to roll out at scale, but they also move part of the risk to the account that syncs them and the device that unlocks them.
Microsoft shared at Authenticate 2026 that 75% of passkeys registered in Microsoft Entra in the previous month were synced. So this is not an edge case anymore. It is quickly becoming the default deployment reality.
The new attacks are against recovery and trust assumptions
The most practical attack in the research is not βbreaking passkeys.β It is hijacking the recovery flow around them.
The UNC6671 phishing kit targets iCloud Keychain recovery. In the demo presented by Arie Olshtein, the victim gets a fake security alert, enters a real Apple verification code relayed by the attacker, and is then tricked into providing the iPhone passcode. That passcode can also act as an iCloud Keychain recovery factor. Once the attacker joins the victimβs keychain, they receive all synced passkeys.
On the endpoint side, Palo Alto Networks described three attack classes against the Google Password Manager cloud authenticator on desktop:
- Pass-ta-Key: signs without real user gesture if the relying party does not strictly require user verification.
- Silver Pass-ta-Key: takes over the mechanism proving user verification.
- Golden Pass-ta-Key: steals the accountβs 32-byte Security Domain Secret and decrypts synced passkeys offline.
Golden Pass-ta-Key is the ugly one. A single secret protects every synced passkey in the Google account, and according to the researchers, Google offers no way to rotate or revoke it today.
The server-side control too many RPs still miss
If there is one technical takeaway here, it is this: setting userVerification: "required" in WebAuthn is not enough.
The relying party has to validate the UV flag in every assertion when policy requires user verification. Palo Alto Networks showed scenarios where Googleβs cloud authenticator still returned valid assertions even when no PIN or biometric was used, but with the UV bit cleared.
That means the browser flow can look fine while the server silently accepts a weaker event than intended.
A simple summary:
| RP assumption | What can go wrong | What to do |
|---|---|---|
userVerification: required guarantees verification |
It only requests verification | Validate the UV flag server-side |
| Passkey means phishing-resistant account security | Fallbacks still get attacked | Remove SMS OTP/password fallbacks where possible |
| A new passkey is always a good sign | Attackers may register one after fallback takeover | Treat post-fallback passkey creation as high risk |
The eBay example is especially useful. The researchers reported that eBay had accepted assertions without validating UV, and eBay now enforces and validates both UV and UP flags, with version-specific waivers where browser behavior is inconsistent.
Most account takeovers still happen through fallbacks
PayPalβs takeaway is blunt: they see no evidence of phishing among passkey logins. The account takeovers still happen through SMS OTP and password fallbacks, and then the attacker registers passkeys in their own credential manager.
That should reset priorities for a lot of product and identity teams. If you keep phishable recovery and fallback paths attached to a passkey account, attackers will just route around WebAuthn.
The controls under RP ownership are pretty clear:
- validate UV and UP server-side
- reduce or remove phishable fallbacks for users who consistently use passkeys
- harden recovery with risk tiers, identity proofing, and cooling-off periods
- alert on a new passkey created right after SMS OTP or password login
For privileged users, device-bound passkeys still deserve more attention because they avoid the sync-account recovery problem entirely. But for consumer-scale adoption, synced passkeys are not optional anymore. The right response is not to avoid them. It is to stop pretending the protocol is the whole system.
Read the full breakdown.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.