Passkeys Raise the Bar. Here’s Where Attackers Can Still Get In.

Passkeys represent a significant advancement in phishing-resistant authentication, yet they do not eliminate account compromise. Attackers have pivoted toward "passkey-themed phishing," where the credential itself is not broken, but used as social engineering bait to steer users toward exploitable authentication flows, such as device code authentication. High-profile campaigns like EvilTokens demonstrate that AI-generated lures can scale these attacks, allowing unauthorized session access without stealing passwords or bypassing cryptography.
True account security requires a layered strategy because authentication is only one part of the account lifecycle. Vulnerabilities persist in enrollment and recovery processes, compromised endpoints, and the storage models of synced versus device-bound keys. Effective defense necessitates integrating phishing-resistant MFA with endpoint detection and response (EDR), identity threat detection and response (ITDR), and continuous security awareness training. The goal is to maintain visibility and containment capabilities after a successful sign-in occurs, ensuring that a valid authentication event does not automatically equate to a legitimate session.

* Huntress researchers tracked the EvilTokens campaign.
* EvilTokens targeted 344 organizations across five countries over 16 days.
* Microsoft researchers documented attacks where passkeys served as lures for non-phishing-resistant MFA flows.
* SpecterOps researched techniques to manipulate Windows WebAuthn interactions via browser hooks.
* CISA identifies phishing-resistant MFA as the gold standard.
* Passkeys are categorized into two storage models: synced and device-bound.
* Device code authentication allows activating apps on a device by entering a code on another.
* FIDO2/WebAuthn hardware keys function as a specific mode of MFA.
* ITDR focuses on identifying and containing suspicious activity surrounding legitimate sign-ins.
* EDR agents monitor managed devices signing into company resources.

The strongest version of this narrative is a cautionary plea for "defense in depth." It correctly identifies that security is a chain and that upgrading one link (authentication) merely shifts the attacker's focus to the next weakest link (enrollment, recovery, or the endpoint).
This is a classic case of a vendor-driven threat intelligence narrative. By framing the "gap" in passkey security, the narrative creates a logical necessity for a suite of complementary products: EDR, ITDR, and specialized training platforms. The evidence for the threat is provided by research firms (Huntress, SpecterOps, Microsoft) to validate the need for the proposed solution ecosystem. While the technical claims are sound, the structure is designed to move the reader from a state of perceived security (passkeys) to a state of recognized vulnerability, which can only be solved by expanding their security stack.
Patterns detected: ARC-0021 Fear Appeal, ARC-0014 Authority Game
The driving paradigm is the "Endless Arms Race." It assumes that no single technology is ever a "solution," only a temporary deterrent. This echoes the historical shift from passwords to MFA, and now from MFA to phishing-resistance. The implication is that human agency is a liability to be managed through "habits" and "training" rather than a component to be empowered. The cost is an ever-increasing complexity of the security stack, which benefits vendors more than the end-user.
If this were an influence campaign, the playbook would be: "Validate the new gold standard $\rightarrow$ Reveal the hidden flaw $\rightarrow$ Position a multi-product suite as the only viable remedy." The content aligns with this structural pattern.
Bridge Questions:
1. If we automate the "trusted enrollment route," does that remove the human vulnerability or simply create a new high-value target for attackers?
2. At what point does the complexity of a "layered defense" create more security gaps (via misconfiguration) than it closes?
3. How would the strategy change if the goal shifted from "detecting compromise" to "minimizing the blast radius" of a successful sign-in?