Adaptive MFA 2026: How Risk-Based Authentication Replaces…

79 percent of business email compromise victims had implemented MFA correctly – and were still hacked. The reason: standard MFA is a one-time hurdle that becomes worthless once tokens are stolen. Adaptive MFA assesses every login attempt in real time and adjusts the security level to the current risk. NIS2 calls for exactly that: “continuous authentication solutions”. What is behind it and how the major providers implement it.

What sets adaptive MFA apart from standard MFA

Standard MFA asks for the same second factor at every login – regardless of whether the user is accessing from their usual office PC or from an unknown IP address in another country. Adaptive MFA, by contrast, evaluates every authentication attempt in real time based on contextual signals and dynamically adjusts the security level.

The signals the system evaluates include device type and status, location and IP reputation, time of day, user behavior, and network context. From this, a machine learning model calculates a risk score that triggers three response levels:

The key difference: standard MFA is a one-time barrier. Once it has been overcome (for example through AiTM token theft), the attacker has free access until the session times out. Adaptive MFA with Continuous Access Evaluation can revoke tokens in real time when the risk profile changes.

Microsoft Entra Conditional Access: The market standard

Microsoft Conditional Access is the most widely used adaptive MFA implementation in the enterprise segment. According to Microsoft, the policy engine of Entra ID (formerly Azure AD) processes more than 40 TB of identity signals per day.

The system calculates two separate risk scores: User Risk (the likelihood that the account has been compromised) and Sign-in Risk (the likelihood that an individual login is unauthorized). Both scores feed into the access decision.

Typical policy configurations in practice:

• Sign-in Risk “Medium” or “High”: enforce MFA, even if the device is known
• Non-compliant device: block access completely, regardless of user risk
• Trusted Location + Managed Device: no MFA required – the combination of network and device status is sufficient as a trust anchor

Okta, Auth0, and Cisco Duo: The Alternatives

Okta Adaptive MFA (from USD 6 per user/month) analyzes devices, network, location, and user behavior. Its proprietary ThreatInsight engine evaluates data from millions of authentications to detect attack patterns. Since Q4 2024, it has also offered Identity Secure Posture Management, which can automatically enforce MFA for critical roles.

Auth0 (part of Okta since 2021, for customer identity scenarios) works with three defined risk signals: NewDevice (login from an unknown device within the last 30 days), ImpossibleTravel (geographically impossible login patterns), and UntrustedIP (suspicious IP from a threat intelligence database). Available only in the Enterprise Plan.

Cisco Duo (in Duo Premier and Advantage) offers Risk-Based Factor Selection and Risk-Based Remembered Devices. A distinctive feature: Wi-Fi fingerprinting via the Duo Desktop App as an additional location signal. Duo also detects Push Harassment and Push Spray attacks as separate risk signals.

“In 65 BEC incidents examined, 79 percent of victims had correctly implemented MFA – and were still compromised because session tokens were stolen.”
– FRSecure Incident Response Report, 2025

NIS2 Requires “Continuous Authentication” – What That Means in Practice

NIS2 Article 21(2)(j) literally refers to: “the use of multi-factor authentication or continuous authentication solutions”. That is more than a recommendation. For the roughly 29,500 affected companies in Germany, it means: standard MFA meets the minimum requirement, but adaptive MFA with continuous evaluation addresses the spirit of the law.

ENISA (the European cybersecurity agency) specifies: MFA is mandatory for privileged accounts, system administration, and access to network and information systems. The personal liability of executive management makes this a board-level issue.

Why Standard MFA Fails Against AiTM Attacks

Adversary-in-the-Middle (AiTM) attacks place a proxy between the user and the login page. The user authenticates correctly – including MFA. The proxy intercepts the resulting session token. From that moment on, the attacker has a valid, authenticated session.

Microsoft reported 147,000 token replay attacks in one year – an increase of 111 percent. The problem: standard MFA is a one-time barrier. After successful authentication, there is no further check until the session times out.

Adaptive MFA with Continuous Access Evaluation (CAE) breaks this pattern. The system continuously checks whether the session still matches the original risk parameters. If the location, device, or behavior pattern changes, the session is invalidated immediately – not hours or days later.

Phishing-resistant MFA (FIDO2/passkeys) goes one step further: the credentials are domain-bound. A proxy cannot forward them because the cryptographic handshake works only with the genuine domain. AiTM attacks become structurally impossible.

Provider comparison: What the major platforms can do

Practical checklist: Introducing adaptive MFA

1. Inventory: Which systems currently use which authentication methods? Which privileged accounts rely only on password + SMS OTP?
2. Define risk policies: For which applications and user groups should step-up authentication apply? At minimum: admin access, cloud consoles, VPN.
3. Enable phishing-resistant factors: FIDO2 security keys or passkeys as the primary factor for privileged accounts. Not just as an option, but as a requirement.
4. Turn on Continuous Access Evaluation: For Entra ID: enable CAE in Conditional Access Policies. For others: reduce token lifetime to the minimum.
5. Configure Impossible Travel: Every platform has thresholds for geographically impossible logins. Enable them and adapt them to realistic travel patterns.
6. Set up monitoring: Feed risk-based login events into the SIEM. Do not just log suspicious patterns, alert on them – the SIT lesson applies here too.

Conclusion: Standard MFA is necessary, but not sufficient

MFA remains the single most important measure against identity attacks – Microsoft puts the effectiveness of phishing-resistant MFA at over 99 percent of all password-based attacks. But standard MFA without risk assessment is vulnerable to token theft and AiTM attacks. Adaptive MFA closes this gap through continuous risk assessment and dynamic security levels.

NIS2 makes the direction binding: “continuous authentication solutions” appears in the legal text. Anyone still working with static MFA today may meet the minimum requirement, but not the spirit of the regulation. The first step: enable Conditional Access or comparable policies for the most privileged 10 percent of accounts. That takes one afternoon of configuration and addresses 80 percent of the risk.

Frequently Asked Questions

Every question is locked. A tap unlocks the answer.