FIDO passkeys and phishing-resistant authentication are resetting IAM

At a glance

What this is: This is an identity and authentication analysis arguing that passkeys and phishing-resistant methods are becoming the baseline for modern access control as phishing, MFA bypass, and AI-assisted fraud intensify.

Why it matters: It matters because IAM teams must now align human authentication, NHI trust, and AI-era access governance around stronger proof of possession and lower reliance on shared secrets.

By the numbers:

• On estime aujourd'hui que le phishing est 30 fois plus rentable que l'ingénierie sociale traditionnelle menée par l'homme.
• Plus de 53 % des consommateurs ont signalé une augmentation des activités numériques suspectes au cours de l'année écoulée.

👉 Read OneSpan's analysis of FIDO passkeys and phishing-resistant authentication

Context

Phishing-resistant authentication is becoming a governance issue, not just an end-user login choice. The article argues that traditional MFA is no longer enough against proxy phishing, real-time adversary-in-the-middle attacks, and AI-assisted social engineering, especially as identity becomes the control point for both human and machine access.

For IAM leaders, the important shift is that authentication design now sits at the intersection of user experience, platform resilience, and AI risk. Passkeys, FIDO standards, and hardware-backed authenticators change how organisations think about assurance, but they also force a broader review of how identity systems are trusted, enrolled, and scaled across different actor types.

Key questions

Q: How should security teams implement phishing-resistant authentication without hurting adoption?

A: Start with the highest-risk populations and applications, then offer the simplest usable authenticators that still meet your assurance target. Build recovery, enrollment, and help desk processes at the same time. If users cannot enroll and recover reliably, they will route around the control and weaken the programme.

Q: When should organisations require hardware-bound keys instead of synchronised passkeys?

A: Require hardware-bound keys for privileged access, regulated operations, and any workflow where origin assurance and device binding matter more than convenience. Synchronised passkeys can work for lower-risk user populations, but they should not be treated as equivalent when the business impact of account compromise is high.

Q: What do security teams get wrong about MFA in phishing-heavy environments?

A: They assume MFA always blocks account takeover, even when the factor can be relayed in real time. Prompt-based or code-based MFA can still fail under proxy phishing and adversary-in-the-middle attacks. The right question is whether the factor is resistant to interception and replay.

Q: How can IAM teams prepare for AI-driven identity fraud?

A: Treat AI fraud as an identity and authorization problem, not only a fraud analytics problem. Strengthen proofing, reduce reliance on reusable secrets, and put stronger controls around recovery, delegated access, and high-risk transactions. That gives fraud teams and IAM teams a shared control model.

Technical breakdown

Phishing-resistant authentication and MFA bypass

Traditional MFA adds a second factor, but it still often depends on a shared secret, a reusable code, or a prompt that can be intercepted in real time. Proxy phishing and adversary-in-the-middle attacks work because they capture the session or relay the challenge instead of cracking the underlying credential. Phishing-resistant methods change the trust model by binding authentication to a cryptographic key pair and origin verification. That removes the reusable secret from the attack path and makes credential replay far harder.

Practical implication: move high-risk access paths away from shared-secret MFA and toward phishing-resistant authenticators.

Passkeys, device binding, and sync trade-offs

Passkeys use public-key cryptography so the private key stays on the device or in a secured sync ecosystem, while the server only stores a public key. That design reduces secret theft and makes authentication easier for users. The trade-off is governance: synchronised passkeys improve usability, but hardware-bound keys generally offer stronger assurance for high-risk roles and regulated workflows. IAM teams need to distinguish between convenience-oriented enrollment and assurance-oriented enrollment, because not every passkey mode is equivalent.

Practical implication: segment passkey policy by assurance tier instead of treating all passkeys as the same control.

Identity and AI: three converging control planes

The article frames three identity relationships with AI: using AI to improve detection and adaptive access, defending identity against AI-generated fraud, and using identity to secure AI models and data. That is the right mental model because AI changes both the attacker side and the control side. Deepfake impersonation and synthetic fraud increase the cost of weak authentication, while AI workloads themselves become subjects of access governance. Identity is no longer only a front-door control; it is also a governance layer for AI use.

Practical implication: align authentication, fraud controls, and AI access governance in one operating model.

Breaches seen in the wild

• Cisco DevHub NHI breach — IntelBroker exploited exposed Cisco credentials, API tokens and keys in DevHub.
• DeepSeek breach — DeepSeek breach exposed 1M+ log lines and sensitive secret keys.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Passkeys are becoming the practical replacement for shared-secret trust in human authentication. The article makes a straightforward point: password-based access and conventional MFA are too easy to relay, prompt, or socially engineer. Phishing-resistant authentication changes the assurance model by removing reusable secrets from the path. For IAM programmes, this is not a cosmetic upgrade, it is a shift from recoverable credentials to cryptographic proof of possession.

Hardware-bound authenticators still matter when assurance is the goal, not just convenience. Synchronised passkeys improve adoption, but they do not create the same risk profile as device-bound security keys. High-assurance use cases still need a tighter binding between identity, device, and origin. That distinction matters for privileged access, regulated workflows, and account recovery design.

AI is now an identity problem as much as an AI problem. The article correctly places AI in three identity lanes: augmenting detection, attacking identity with synthetic fraud, and requiring identity controls around models and data. That framing is important because it ties human IAM, fraud resistance, and emerging AI governance into one control narrative. Practitioners should stop treating AI fraud and identity assurance as separate workstreams.

Authentication experience is now a deployment constraint, not an afterthought. The article emphasises usability, ecosystem simplicity, and authenticator choice because weak adoption is itself a security failure mode. If the secure path is painful, users route around it and the programme fails at scale. The implication for practitioners is clear: assurance without workable enrollment and recovery does not hold in production.

Phishing-resistant authentication is now the minimum viable control for modern access governance. As phishing economics improve and AI lowers attacker effort, organisations that still rely on promptable MFA are operating on a fading assumption. The implication is not simply to add another factor, but to redesign the authentication trust model around origin-bound, non-relayable credentials.

From our research:

• The attacks on phishing have increased by 4.2% after the introduction of ChatGPT, according to LLMjacking: How Attackers Hijack AI Using Compromised NHIs.
• In the same research stream, attackers attempt access to exposed AWS credentials in an average of 17 minutes, which shows how quickly compromised secrets become operational risk.
• For a broader view of how AI changes NHI exposure, see Ultimate Guide to NHIs , 2025 Outlook and Predictions for the identity governance implications of agentic and machine access.

What this signals

Phishing-resistant authentication is now a programme design issue, not a point solution choice. Teams that still rely on promptable MFA are carrying forward an assumption that attackers need longer than they actually do. The practical signal is to segment access by assurance tier and prioritise systems where session relay or push fatigue would create the biggest blast radius.

Passkey policy will increasingly split along assurance lines. The governance decision is no longer whether to adopt passkeys, but which form of passkey is acceptable for which identity and risk class. That means enrollment, recovery, and privileged access policy must be written together, not as separate workstreams.

Identity teams should expect AI to increase pressure on both authentication and recovery paths. Deepfake-assisted impersonation raises the value of origin-bound credentials, while automated social engineering makes fallback channels more attractive to attackers. Organisations that tighten recovery controls and reduce relayable factors will have a materially better operating position.

For practitioners

• Prioritise phishing-resistant authentication for high-risk access paths Start with privileged users, finance, admin consoles, and any workflow that can trigger sensitive changes. Replace promptable MFA where attacker-in-the-middle risk is highest, then expand to broader workforce access after enrollment and recovery are stable.
• Separate convenience passkeys from assurance-grade authenticators Define which user groups can use synchronised passkeys and which roles require hardware-bound keys. Document the decision by application tier, data sensitivity, and recovery requirements so the policy is enforceable instead of aspirational.
• Rework account recovery as an assurance control Treat recovery as part of the authentication boundary, not an operational exception. Require strong identity proofing, resistant backup methods, and explicit recovery logging because many phishing-resistant programmes fail at the fallback path.
• Extend identity governance to AI access cases Map where AI systems, models, or agents need access to sensitive data and enforce the same authorization discipline used for human identities. Separate model use, data use, and administrative use so access does not blur into standing trust.
• Measure where users still depend on relayable authentication Inventory login methods that still allow push fatigue, OTP interception, or session relay. Use that baseline to target the top applications and the highest-risk user groups first, then track migration by application and identity type.

Key takeaways

• Phishing-resistant authentication is replacing promptable MFA as the practical baseline for modern identity assurance.
• Passkeys improve usability, but hardware-bound authenticators still matter where origin assurance and device binding are essential.
• AI expands identity risk across login, recovery, and access governance, so IAM programmes must treat authentication as part of a wider control model.

Key terms

• Phishing-resistant authentication: Authentication designed to resist interception, replay, and real-time relay attacks. It usually relies on cryptographic proof of possession and origin binding rather than a reusable secret or one-time code. In practice, it raises the cost of account takeover because the attacker cannot simply forward the login challenge.
• Passkey: A passkey is a credential based on public-key cryptography that replaces a password with a key pair. The private key remains on a device or in a secure sync system, while the service stores only the public key. In identity programmes, the assurance level depends on how the passkey is stored, recovered, and bound to the user or device.
• Hardware-bound authenticator: A hardware-bound authenticator is a physical device that stores or uses the private key locally and ties login to possession of that device. It is often used for higher-assurance access because it is harder to clone, intercept, or relay than a shared-secret factor. The control value comes from the device binding, not just the presence of a second factor.
• Adversary-in-the-middle attack: An adversary-in-the-middle attack relays authentication traffic between a user and a legitimate service so the attacker can capture the session or replay it. The technique can defeat weaker MFA methods because the attacker does not need to crack the credential, only sit in the communication path long enough to hijack the exchange.

Deepen your knowledge

Phishing-resistant authentication and passkey strategy are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are redesigning access assurance across human, workload, and AI-facing identities, it is a strong starting point.

This post draws on content published by OneSpan: Retour à l'identité, l'alliance FIDO et l'avenir de l'authentification résistante à l'hameçonnage. Read the original.