The Passkey Ecosystem: Attacks Beyond FIDO2 Cryptography
Passkeys: The Shifting Attack Surface Beyond Cryptography
Passkeys have rapidly emerged as a cornerstone of modern, phishing-resistant authentication. Their foundation in FIDO2 public-key cryptography promises a robust defense against common credential-based attacks. However, recent disclosures from security researchers paint a critical picture: while the underlying cryptography remains sound, the broader passkey ecosystem—encompassing enrollment, device management, recovery, and endpoint security—introduces a new set of vulnerabilities that enterprise security teams cannot ignore.
The Illusion of Cryptographic Infallibility
The initial narrative surrounding passkeys often emphasized their cryptographic strength, suggesting that by replacing passwords with public-key pairs, the vast majority of phishing and credential theft risks would be eradicated. This is fundamentally true at the protocol level. A passkey, by design, binds the credential to the legitimate service and keeps the private key off the server, making traditional phishing attacks incredibly difficult. However, the ecosystem around passkeys is proving to be a fertile ground for new attack vectors.
Recent research, including findings from SpecterOps, has identified at least 39 distinct methods to compromise passkey-protected accounts without ever breaking the FIDO2 cryptographic protocol itself [7, 10]. These methods exploit weaknesses not in the cryptography, but in the implementation, integration, and operational layers that support passkeys. This distinction is critical for enterprise security architects and IAM engineers: the threat is no longer solely about cracking the cryptographic algorithm, but about manipulating the environment in which passkeys operate.
Where Attacks Are Manifesting: The Ecosystem's Weak Points
The identified attack methods broadly fall into categories that expose the vulnerabilities of the passkey ecosystem rather than the core FIDO2 protocol. Understanding these areas is paramount for developing a comprehensive defense strategy.
1. Social Engineering and Impersonation
Despite passkeys' resistance to phishing, social engineering remains a potent threat. Threat actors are adapting their tactics to leverage the context of passkey adoption. Microsoft Security Research and OffSeq have detailed active cloud-based intrusions leveraging passkey-themed social engineering. Attackers impersonate IT helpdesk personnel via phone or SMS, luring victims to sophisticated phishing sites designed to mimic legitimate sign-in pages [6, 9].
These attacks often employ adversary-in-the-middle (AiTM) techniques or device code authentication flows to harvest credentials. Once compromised, threat actors establish persistence by adding unauthorized multi-factor authentication (MFA) methods and conducting extensive reconnaissance using APIs like Microsoft Graph. This ultimately leads to large-scale data exfiltration from cloud services such as SharePoint, OneDrive, and Exchange. The critical insight here is that the user is tricked into authorizing an action that bypasses the intended security of the passkey, rather than having their passkey directly stolen.
2. Endpoint Compromise and OS Vulnerabilities
Passkeys reside on devices—smartphones, laptops, and security keys. The security of the passkey is inextricably linked to the security of the host device. If an endpoint is compromised, an attacker can manipulate the authentication process, even if the passkey itself isn't directly extracted.
SpecterOps research, specifically demonstrated at Black Hat USA, revealed how vulnerabilities in operating systems could be chained with other flaws to bypass phishing-resistant MFA. For instance, an issue with how Windows stored old YubiKey signatures in plain text, combined with a flaw in Entra ID's passkey checks, allowed an attacker to impersonate a privileged account [8]. Microsoft has since released a fix (CVE-2026-34348), highlighting the ongoing need for rigorous patching and endpoint security.
Malware designed to target passkey implementations can manipulate the authentication flow without requiring a cryptographic break. This includes scenarios where an attacker with control over the device can initiate an authentication request and approve it, effectively using the legitimate user's passkey without explicit consent or knowledge. The