Actiknow
Custom Software Development

Custom Software Multi-Factor Authentication: When MFA Should Be Mandatory and for Whom

Design a risk-based MFA policy for custom B2B software. Learn who needs mandatory MFA, when to require step-up verification, and how to secure recovery.

Cybersecurity architect presenting a multi factor authentication strategy for a custom b2b application to business executives
Contents hide

Executive summary

Multi-factor authentication (MFA) is often implemented as a checkbox: switch it on for everyone, or leave it optional to avoid complaints. Both approaches miss the central design question. Which identities, transactions, and recovery paths could cause meaningful damage if a password or session is compromised? The answer should determine where MFA is mandatory, what assurance is required, and when a user must authenticate again.

For most business applications, administrators and privileged operators should use mandatory, preferably phishing-resistant MFA. Users who can approve payments, change bank details, export sensitive records, impersonate customers, or modify security settings should face stronger verification before those actions. Routine low-risk activity may justify fewer prompts, but only after the team has considered data sensitivity, account takeover exposure, regulatory obligations, and the cost of an unauthorized action.

MFA does not replace authorization, tenant isolation, secure sessions, or audit logs. It is one layer in an identity architecture. A well-designed program makes strong verification predictable for risky work while keeping everyday tasks usable.

1. Start with business consequences, not login-screen preferences

The real unit of analysis is not simply the user. It is the combination of identity, privilege, resource, action, and context. A support agent reading a ticket and the same agent exporting an entire customer database should not necessarily receive the same authentication treatment. Likewise, a customer browsing public product information and an account owner changing payout instructions have different risk profiles.

Begin with an inventory of business-impacting events: account takeover, fraudulent approval, data extraction, tenant-wide configuration changes, loss of administrative access, and unauthorized API credential creation. Record which roles can initiate each event and whether the application already has preventive controls such as dual approval, transaction limits, or separate authorization.

Product owner and cybersecurity analyst assessing user permissions and business risks for a custom application

For custom products, this exercise belongs in the architecture and product requirements, not as a late security add-on. Actiknow’s custom web application development services describe the broader application lifecycle, including design, development, testing, deployment, and ongoing support. Authentication decisions affect each of those stages.

2. Who should be required to use MFA?

Mandatory by default: administrators and privileged identities

Require MFA for platform administrators, tenant administrators, security administrators, database and infrastructure consoles, billing administrators, and staff who can create or elevate other users. Favor phishing-resistant methods for these roles. A compromised admin account can override ordinary application controls, so a weak fallback method can undermine the entire policy.

Do not exempt internal employees because they connect from an office network or VPN. Network location is a risk signal, not a second authentication factor. Shared administrator accounts should be eliminated where possible; individual identity, least privilege, and attributable audit events are more useful than a shared credential with an MFA code.

Mandatory for sensitive business workflows

Consider MFA mandatory for people who regularly handle regulated or high-impact records, financial approvals, payroll changes, refunds, beneficiary updates, patient information, confidential client files, and bulk exports. Depending on the workflow, require it at sign-in, at the sensitive action, or both. For example, a finance approver who uses the application throughout the day may authenticate strongly at login and then perform a fresh verification when approving an unusually large transfer.

Risk-dependent for ordinary business users

A read-only employee who views non-sensitive operational metrics may not need a prompt on every session. Yet optional MFA should be an explicit risk decision, not an unexamined default. If the account exposes customer data, can access multiple tenants, or can initiate costly actions, the baseline should rise. Where enterprise customers manage identities through SSO, the product should support customer-defined enforcement without creating an unprotected local-login bypass.

External customers, partners, and occasional users

For external users, choose a policy based on the consequences of compromise and the friction of onboarding. A low-risk community portal may encourage MFA and require it only for account recovery or changes to sensitive details. A healthcare, financial, or partner-management portal may need stronger authentication from the outset. Accessibility, device availability, and recovery support are product requirements, not reasons to leave privileged workflows unprotected.

Service accounts and automated integrations

Non-human identities cannot complete interactive MFA in the ordinary sense. Do not make a robot share a human’s account to bypass MFA. Use workload identities, short-lived credentials, certificate or key-based authentication, managed secrets, narrow scopes, and rotation. Review whether each integration needs the permissions it has. Human approval or privileged configuration changes to those integrations should still be protected by MFA.

3. Make an MFA policy matrix that product and security teams can operate

A useful policy is concise enough for a product manager to explain and precise enough for an engineer to enforce. The following example is illustrative, not a universal compliance standard.

Platform administrator | Mandatory at sign-in | Phishing-resistant method preferred or required by policy | Fresh verification for security-policy changes, privilege grants, and emergency access.

Tenant administrator | Mandatory at sign-in | Strong method | Fresh verification for user invitations with elevated roles, SSO changes, API token issuance, and tenant-wide exports.

Finance or sensitive-data operator | Mandatory for access to protected functions | Strong method | Step-up for payment approvals, bank-account changes, high-value refunds, and bulk downloads.

Ordinary employee | Mandatory or conditional based on data classification | Organization-approved method | Step-up when entering a high-risk workflow or unusual context.

Low-risk external user | Encourage or require according to account risk | Usable approved method | Step-up for credential changes, sensitive profile changes, and high-impact transactions.

Workload identity | No interactive MFA | Managed non-human credentials | Separate privileged approval for changes to scopes, keys, and integration configuration.

For each row, assign a policy owner, permitted authenticators, exception process, evidence of enforcement, and review date. A matrix without operational ownership quickly becomes an outdated spreadsheet.

4. MFA is not one technology: choose the right assurance

Passkeys and security keys

FIDO2/WebAuthn-based authentication can resist phishing by binding credentials to the legitimate relying-party origin. When a passkey requires user verification, such as a device PIN or biometric check, it can satisfy multi-factor requirements through a cryptographic authenticator. For high-privilege roles, organizations may prefer hardware-backed, non-exportable credentials and carefully controlled recovery. Account for supported devices, browsers, shared-workstation policies, and backup authenticators.

Security professional testing passkey authentication and a hardware security key for a business application

Authenticator-app one-time codes

Time-based one-time passwords (TOTP) are widely deployable and generally preferable to relying on SMS for routine MFA. They remain vulnerable to real-time phishing or code relay, so do not treat them as equivalent to origin-bound authentication. Protect enrollment seeds, avoid logging codes, limit attempts, and require secure factor replacement.

Push authentication

Push can be convenient but can introduce approval fatigue if users receive repeated unexpected prompts. Use challenge or number matching where supported, cap repeated requests, and provide a clear way to report suspicious prompts. A simple Approve button without transaction context is a poor control for high-impact operations.

SMS and email

SMS is susceptible to SIM swapping, interception, and delivery failures; it may be a constrained fallback where policy permits but should not be the preferred option for high-risk accounts. Email-based confirmation can be useful for notifications or account verification, but an emailed code should not be casually labeled a strong, independent second factor when the email account is itself exposed through the same compromised session or device.

The NIST digital identity guidelines distinguish authenticator assurance levels and phishing resistance. The OWASP multifactor authentication guidance also emphasizes independent factors, risk-based prompts, secure recovery, and defenses against MFA bypass. Use these as design inputs, then validate any sector-specific legal or contractual requirements separately.

5. SSO does not automatically mean MFA is enforced

Enterprise SSO can simplify sign-in and centralize identity policies, but federation is not proof that a particular user completed MFA. A custom application should define which identity providers are trusted, which tenants can configure federation, and which authentication-context claims or session signals it requires for protected operations.

For OpenID Connect or SAML integrations, evaluate the identity provider’s MFA policy, authentication context, authentication time, token audience, issuer, and signature. Do not blindly trust a front-end flag such as mfa=true. Where the identity provider cannot convey sufficient assurance, the application may need a supported step-up flow or an explicit contractual control rather than silently assuming the requirement was met.

Plan for local account recovery, invitation acceptance, emergency access, and federation outages. A strong SSO policy is ineffective if a forgotten-password screen creates a password-only route to the same privileged account. Likewise, when a tenant disables SSO, the product needs an approved transition and security review.

6. Require step-up MFA for the actions that matter most

Step-up authentication asks for fresh or stronger proof when a user attempts a sensitive operation. It should be enforced on the server after checking authorization and before the business transaction commits. A modal in the browser alone is not a security boundary.

Common triggers include adding an administrator; modifying role or permission policies; exporting large datasets; changing bank or payroll details; issuing API credentials; resetting another user’s MFA; impersonating a customer; disabling audit logging; and modifying SSO configuration. For particularly sensitive actions, add maker-checker approval, transaction limits, or a delay and out-of-band notification.

Finance manager and security professional reviewing additional authentication for a sensitive payment approval

A reliable flow is: the user requests the action; the backend checks permission and risk; the server determines whether the session has adequate recent authentication; the user completes the approved challenge; the backend validates the assurance and freshness; the original action is re-authorized and executed once; the outcome is audited. Use a short-lived, purpose-bound step-up result, not a permanent account-wide bypass.

Illustrative policy logic: require_step_up = privileged_action OR sensitive_export OR high_value_transaction OR suspicious_session. This expression is a starting point for a policy engine, not production-ready code. Define action names, thresholds, trusted signals, freshness windows, and exception handling explicitly.

7. Treat recovery and MFA enrollment as protected workflows

Many MFA programs are defeated through the help desk rather than the login page. Attackers may try to replace an authenticator, reset a phone number, add a recovery email, or persuade support staff to remove MFA. These paths should have at least as much attention as the primary sign-in experience.

At enrollment, require an authenticated session, validate possession of the new factor, record factor type and enrollment time, and notify the account owner. For privileged users, consider registering a backup phishing-resistant authenticator before activation. Recovery codes should be single-use, securely stored, and presented with clear instructions.

For factor replacement, require an existing factor where available; otherwise use a documented recovery process with identity verification appropriate to the account’s risk. High-privilege recovery may need dual approval, delayed activation, temporary restrictions, and a security notification. Support agents should not be able to both approve their own bypass and suppress the audit trail.

It support specialist assisting an employee with secure mfa recovery using a backup security key

Test realistic failure cases: lost phone, changed employer, inaccessible email, broken security key, traveling employee, disabled identity provider, and a legitimate admin locked out during an incident. Recovery is part of the user journey, not a footnote in the security policy.

8. Protect sessions after MFA succeeds

MFA establishes confidence at an authentication event; it does not guarantee that a session remains safe. Stolen session cookies, token leakage, malware, cross-site scripting, and misconfigured refresh tokens can still lead to compromise. Enforce secure cookie attributes, TLS, sensible session lifetime, token rotation where appropriate, server-side revocation, and reauthentication after material risk events.

For a web application, prefer HTTP-only secure cookies for browser sessions where architecture allows, and defend against CSRF as appropriate. For mobile applications, use platform-protected storage and approved authentication libraries. Log session creation, high-risk changes, revocation, and authentication failures without recording passwords, OTPs, recovery codes, or raw bearer tokens.

A privileged session may warrant a shorter reauthentication interval than a routine read-only session. Choose values based on risk and applicable standards, then test the effect on real workflows. Avoid the opposite extremes of a prompt on every click and a privileged session that remains valid indefinitely.

9. Design MFA for multi-tenant B2B applications

Multi-tenant software adds an extra policy boundary: one customer’s identity choices must not weaken another customer’s controls. Keep platform-wide minimums separate from tenant-specific settings. A tenant may strengthen its requirements, but should not be allowed to disable mandatory controls for platform support staff or privileged application functions.

Map the same human to the correct tenant membership and role at authorization time. Do not infer tenant access solely from an email domain. An identity can be a read-only member of one tenant and an administrator of another. Step-up decisions should use the active tenant and the requested action, not just a global user flag.

Provide administrators with an understandable configuration screen: enforced factors, SSO requirements, trusted identity provider, recovery contacts, break-glass procedure, policy changes, and audit history. If a customer requires phishing-resistant MFA for all privileged roles, expose enough evidence to demonstrate the rule is actually applied.

10. Implementation architecture: where enforcement belongs

A practical design separates identity proofing, authentication, authorization, business action enforcement, and audit. An identity provider or managed authentication service may handle factor enrollment and challenges. The application owns business-role mapping and decides whether the current session meets the required assurance for each action. Backend endpoints must make the final decision.

Store the minimum necessary assurance metadata: user ID, tenant ID, identity provider, authenticated-at time, methods or assurance context, factor policy version, and the step-up event reference. Avoid storing secrets in application logs. For API calls, validate token issuer, audience, expiry, scopes, and tenant context; never infer MFA solely from the presence of an access token.

Where teams are modernizing a legacy application, phase the change. First centralize authentication entry points and inventory bypasses. Then introduce policy enforcement for privileged roles, protect sensitive actions, and retire legacy login paths. Actiknow’s custom solutions offering covers web and mobile applications and integrations, which are the components that typically need to be aligned during such a change.

11. A realistic implementation example: a B2B operations portal

Consider a hypothetical multi-tenant operations portal with customer administrators, regional managers, field staff, and a small internal support team. Field staff submit routine reports. Regional managers can approve adjustments. Tenant administrators can invite users and export records. Internal support can troubleshoot accounts but should not routinely view customer data.

A defensible policy would require phishing-resistant MFA for internal platform administrators, mandatory MFA for tenant administrators, and step-up for large exports and role changes. Regional managers would receive fresh verification for high-value approvals. Field staff could use an appropriately secured session for routine submissions, with stronger verification for credential or recovery changes. Support impersonation would require explicit authorization, a reason, time limit, and audit record.

The exact rules would change if the portal handled medical records, financial transfers, or sensitive minors’ information. The point is not that one MFA matrix fits every industry. It is that the team can explain each requirement in terms of business consequences and verify that the application enforces it.

A published Actiknow telehealth application case study describes the use of multi-factor authentication alongside role-based access controls and secure storage for sensitive healthcare workflows. It is a relevant illustration of the type of application context in which authentication must be considered with broader security design, not evidence that any one MFA configuration is universally sufficient.

12. Measure adoption, enforcement, and friction together

Executives need more than the percentage of users enrolled in MFA. An account may have a factor registered yet still reach a privileged function through an unprotected legacy endpoint. Measure enforcement coverage at the point of action.

Recommended dashboard measures include: percentage of privileged accounts with an approved MFA method; percentage of privileged sign-ins satisfying policy; protected-action step-up success and bypass counts; factor-enrollment completion; recovery and factor-reset volume; suspicious prompt and failed-challenge trends; time to revoke compromised sessions; and customer support contacts related to authentication.

Security operations analyst and it manager monitoring mfa enforcement, authentication compliance and access audit metrics

Segment by tenant, role, application version, factor type, and critical workflow. Investigate spikes in abandonment after enrollment, but do not weaken security silently to improve a conversion metric. An effective dashboard pairs security outcomes with operational burden, such as time to complete a protected approval and the percentage of legitimate users unable to recover access.

13. A six-week rollout plan for an existing custom application

Week 1: Map identities, privileges, and bypasses

Inventory login paths, SSO tenants, local credentials, service accounts, admin roles, sensitive actions, session behavior, and recovery mechanisms. Agree on data classification and risk owners. Identify endpoints that currently trust client-side role checks.

Week 2: Write the policy and select authenticators

Define mandatory roles, factor requirements, step-up triggers, permitted fallbacks, exception approval, recovery rules, and audit evidence. Test device support and accessibility. Decide whether an existing identity provider can enforce the required assurance or whether application changes are necessary.

Week 3: Implement server-side controls

Build or configure factor enrollment, authentication-context validation, protected-action checks, recovery safeguards, and audit events. Include test cases for expired assurance, session theft, replayed requests, missing claims, tenant switching, and revoked users.

Week 4: Pilot with privileged staff and one customer cohort

Start with a limited population, provide backup-factor enrollment, and run recovery drills. Review false challenges, failed sign-ins, user confusion, and help-desk handling. Test security controls through direct API calls, not only the browser interface.

Week 5: Enforce high-risk workflows

Roll out mandatory MFA to remaining privileged identities and sensitive transactions. Disable legacy authentication routes that evade the policy. Communicate the change in advance, provide concise instructions, and monitor exceptions daily.

Week 6: Audit, measure, and stabilize

Verify enforcement against the original role-action matrix. Review recovery and break-glass events, sample high-risk audit records, test offboarding, and assign ownership for ongoing policy reviews. Do not declare success solely because enrollment numbers are high.

14. Common implementation mistakes

Mistake 1: Treating an OTP plus a password as phishing-resistant MFA. A code can be relayed by a convincing phishing site. Match the authenticator to the threat.

Mistake 2: Enforcing MFA only in the front end. An attacker can call the API directly unless the backend checks authentication assurance.

Mistake 3: Protecting sign-in but not recovery. An email-only reset or unreviewed help-desk bypass can defeat a strong login requirement.

Mistake 4: Assuming SSO guarantees MFA. Verify the identity provider’s policy and the evidence your application actually receives.

Mistake 5: Treating a trusted IP address or known device as a second factor. Such signals can influence risk, but they do not themselves prove possession of an independent authenticator.

Mistake 6: Applying identical prompts to every task. Excessive friction encourages workarounds; risk-based step-up and appropriately managed sessions are more sustainable.

Mistake 7: Ignoring service accounts. Automation needs strong non-human identity controls, not password-sharing or an MFA exemption for an overprivileged user.

Mistake 8: Measuring enrollment rather than actual protection. Test that sensitive endpoints reject insufficient authentication, including after token refresh, tenant switches, and factor resets.

15. Executive decision checklist

Before approving an MFA release, ask: Have we named every privileged role? Which actions can cause irreversible financial, privacy, or operational harm? Can those actions be reached through APIs or legacy logins without step-up? Do we know which authentication methods are phishing-resistant? Can customers enforce stricter tenant policies? How are factor resets approved and audited? Have we tested lost-device and SSO-outage scenarios? Do we have measurable enforcement evidence, not just enrollment statistics?

If the team cannot answer these questions, the work is not finished even if the login screen displays a second-factor prompt.

Frequently asked questions

1. Should MFA be mandatory for every user of a custom B2B application?

Not automatically in every low-risk product, but mandatory protection is usually appropriate for privileged identities and sensitive workflows. Decide the baseline from data exposure, transaction risk, customer obligations, and available recovery support. Where the consequences of account takeover are significant, broader mandatory MFA is justified.

2. Should administrators always use phishing-resistant MFA?

For high-privilege accounts, phishing-resistant authenticators are the preferred direction and may be required by the organization’s policy or applicable standards. Hardware-backed credentials can provide additional protection when the operational model supports them.

3. Is an authenticator app enough for finance approvals?

TOTP adds protection over a password but is not phishing-resistant. For high-impact approvals, evaluate origin-bound authentication, fresh step-up, transaction context, limits, and dual authorization rather than relying on a code alone.

4. Can we rely on our enterprise SSO provider?

Yes, if its policy, configuration, and transmitted authentication context meet your requirements. Validate claims and prevent local-login or recovery bypasses. SSO integration is not a substitute for application authorization.

5. What is step-up authentication?

It is fresh or stronger verification requested for a sensitive action or elevated risk. The server must verify the assurance before committing the operation.

6. Is a biometric prompt automatically MFA?

No. A biometric used only to unlock a local device or app may not provide independently verified multi-factor authentication to the server. Properly implemented passkeys with user verification can provide multi-factor cryptographic authentication; the architecture matters.

7. Should we allow SMS as a backup?

Only after considering risk, threat model, contractual obligations, and better alternatives. SMS is weaker against SIM swaps and interception. Avoid making a weak backup the easiest path into privileged accounts.

8. How should users recover access after losing their phone?

Offer pre-enrolled backup factors or one-time recovery codes where appropriate. For privileged accounts, use stronger identity verification, approval, notifications, and auditable controls. Recovery should not quietly remove MFA.

9. Do service accounts need MFA?

They need strong non-interactive authentication rather than human prompts. Use workload identities, scoped credentials, rotation, and monitoring. Protect the human actions that create or modify those credentials.

10. Can a trusted office network replace MFA?

No. Location is a useful signal for conditional access but not an independent authentication factor. Attackers can compromise devices and accounts inside trusted networks.

11. What should we measure after rollout?

Track policy-compliant privileged sign-ins, protected-action enforcement, recovery events, failed challenges, exceptions, user friction, and legacy bypass attempts. Enrollment alone is insufficient.

12. How long does an MFA implementation take?

It depends on identity providers, legacy login paths, number of tenants, sensitive workflows, and recovery complexity. A six-week plan can organize a focused rollout, but it is an illustrative sequence, not a guaranteed delivery estimate.

The bottom line

The strongest MFA policy is not the one that generates the most prompts. It is the one that makes high-risk actions hard to hijack, makes recovery safe, and produces evidence that the controls actually work. Start with privileged identities and sensitive transactions, build server-side enforcement, and test every alternative path into the application.

Discuss an authentication design for your application

If you are planning a new B2B application or revisiting authentication in an existing product, Actiknow can help review the workflow, identity integration points, and implementation requirements as part of its custom application development work. Share your application context through the Actiknow contact page to discuss scope and next steps.

Further reading

NIST SP 800-63B-4: Authentication and Authenticator Management. The guideline explains authenticator assurance, phishing resistance, lifecycle management, and reauthentication.

OWASP Multifactor Authentication Cheat Sheet. Practical guidance on MFA choices, risk-based authentication, factor changes, recovery, and common bypasses.