Customize Adaptive MFA

You can customize Adaptive MFA for a variety of scenarios with Auth0 Actions.

If your users are not enrolled in MFA, you should use the default policy for Adaptive MFA. If a user is not enrolled in MFA and your Action assesses a high risk, you have limited options to stop a bad actor.

Before you begin to customize Adaptive MFA, ask yourself a few questions:

• At what confidence level do you want to trigger MFA?

• How do you want to measure risk?

• Do you want Auth0 to measure confidence or do you want a custom measurement?

• How will you handle users who are not enrolled in MFA?

Adaptive MFA calculates an overall confidence score based on the analysis of three assessments. Each assessment has its own confidence score. To learn more, read Adaptive MFA.

Confidence scores and their associated actions are described below:

If you want to implement your own method for evaluating the overall confidence score of different scenarios, you can use the data available in the riskAssessment object.

Read the examples below to learn how Adaptive MFA would score the confidence of different use cases.

The following table describes high-risk scenarios that result in a low confidence score:

The following table describes low-risk scenarios that result in a high confidence score:

The riskAssessment object contains the overall confidence score, versioning information, and details of the individual assessments.

exports.onExecutePostLogin = async (event, api) => {
  if (event.authentication && event.authentication.riskAssessment) {
    event.authentication.riskAssessment = {
      confidence: 'low' | 'medium' | 'high' | 'neutral',
      version: '1',
      assessments: {
        UntrustedIP: {
          confidence: 'low' | 'medium' | 'high' | 'neutral',
          code: 'not_found_on_deny_list' | 'found_on_deny_list',
          details: { // only if 'found_on_deny_list'
            ip: '192.168.1.1',
            matches: '192.168.0/64',
            source: 'firehol',
            category: 'abuse'
          }
        },
        NewDevice: {
          confidence: 'low' | 'medium' | 'high' | 'neutral',
          code: 'match' | 'partial_match' | 'no_match',
          details: {
            device: 'known' | 'unknown',
            useragent: 'known' | 'unknown',
          }
        },
        ImpossibleTravel: {
          confidence: 'low' | 'medium' | 'high' | 'neutral',
          code: 'missing_geoip', | 'anonymous_proxy' | 'unknown_location' | 'initial_login' | 'location_history_not_found' | 'invalid_travel' | 'minimal_travel_from_last_login' | 'impossible_travel_from_last_login' | 'substantial_travel_from_last_login' | 'travel_from_last_login'
        } 
      },
       PhoneNumber: {
          code: "requires_verification | ok",
          confidence: "low | medium | high | neutral",
          details: {
	        	lineType: "FIXED_LINE | MOBILE | FIXED_LINE_OR_MOBILE | TOLL_FREE | PREMIUM_RATE | SHARED_COST | VOIP | PERSONAL_NUMBER | PAGER | UAN | UNKNOWN"
	            isValid: true | false,
		        countryCode: 1,
		        number: "+12223334444"
        }
      }
    };
  }
}

The assessments object contains details of the three individual risk assessments:

1. NewDevice assessment

2. ImpossibleTravel assessment

3. UntrustedIP assessment  

4. PhoneNumber assessment

Each assessment includes a confidence score, a code that describes the evaluation result, and additional contextual information.

The NewDevice assessment determines if the user is logging in from a known device. It contains the following properties:

If the code equals match, partial_match, or no_match, the details object is present and contains the following properties:

The ImpossibleTravel assessment determines if the user is logging in from a location that would indicate impossible travel. It contains the following properties:

The UntrustedIP assessment determines if the user’s IP address is present in Auth0’s repository of low-reputation IP addresses (“deny list”). It contains the following properties:

If the code equals found_on_deny_list, the details object is present and contains the following properties:

The category property describes the general reason why Adaptive MFA considers a given IP address untrusted. The following table shows all possible values and their descriptions:

The PhoneNumber assessment assesses the risk of a phone number for an incoming transaction. It contains the following properties:

The details object for the PhoneNumber assessment contains the following properties:

If any of your Actions trigger MFA based on confidence score, the default Adaptive MFA policy triggers MFA when the confidence score is low.

The following table shows the possible outcomes based on the combination of Actions and default Adaptive MFA policy actions.

Auth0 provides two Action templates based on Adaptive MFA for you to customize: Adaptive MFA and Require MFA Enrollment.

This template provides an example and starting point for how to build a custom business flow using individual risk assessments.

/**
* Handler that will be called during the execution of a PostLogin flow.
*
* @param {Event} event - Details about the user and the context in which they are logging in.
* @param {PostLoginAPI} api - Interface whose methods can be used to change the behavior of the login.
*/
exports.onExecutePostLogin = async (event, api) => {
  if (event.authentication && 
      event.authentication.riskAssessment && 
      event.authentication.riskAssessment.assessments.NewDevice) {

  // Example condition: prompt MFA only based on the NewDevice 
    // confidence level, this will prompt for MFA when a user is logging in 
    // from an unknown device.
    let shouldPromptMfa;

    switch (event.authentication.riskAssessment.assessments.NewDevice.confidence) {
      case 'low':
      case 'medium':
        shouldPromptMfa = true;
        break;
      case 'high':
        shouldPromptMfa = false;
        break;
      case 'neutral':
        // When this assessor has no useful information about the confidence, 
        // do not prompt MFA.
        shouldPromptMfa = false;
        break;
    }

      // It only makes sense to prompt for MFA when the user has at least one 
      // enrolled MFA factor.
    const canPromptMfa = event.user.multifactor && event.user.multifactor.length > 0;

    if (shouldPromptMfa && canPromptMfa) {
      api.multifactor.enable('any', { allowRememberBrowser: true });
    }
  }
};

This template demonstrates how you could enforce MFA enrollment when using a standard or Adaptive MFA policy. It uses event.user.multifactor to check if the user is enrolled in MFA, and if they’re not, prompts for enrollment.

/**
* Handler that will be called during the execution of a PostLogin flow.
*
* @param {Event} event - Details about the user and the context in which they are logging in.
* @param {PostLoginAPI} api - Interface whose methods can be used to change the behavior of the login.
*/
exports.onExecutePostLogin = async (event, api) => {
  if (!event.user.multifactor || event.user.multifactor.length == 0) {
    api.multifactor.enable('any', { allowRememberBrowser: true });
  }
};

Here are some suggestions for how to build custom Actions based on your use case.

Assess the riskAssessment.confidence property, and then compare it with the constants high, medium, or low:

exports.onExecutePostLogin = async (event, api) => {
  const { riskAssessment } = event.authentication || {};
  const riskIsMedium = riskAssessment && riskAssessment.confidence === 'medium';

  if (riskIsMedium) {
    // ....
  }
}

Confidence scores are discrete values—not in a range—so you cannot use comparison operators (such as < or >) to evaluate multiple values in a single condition.

Use multiple conditions to logically combine all the confidence scores you want to handle. For example, if you want to know when the confidence score is greater than low, check if it’s equal to medium or high:

exports.onExecutePostLogin = async (event, api) => {
  const { riskAssessment } = event.authentication || {};
  const riskIsMediumOrHigh = riskAssessment && 
                                  (riskAssessment.confidence === 'high' || 
                                   riskAssessment.confidence === 'medium');

  if (riskIsMediumOrHigh) {
    // ...
  }
}

The riskAssessment object is saved in your tenant logs. You can view log entries to see the risk assessment score and the determining factors (reasons).

You can view the riskAssessment object and report the results elsewhere. For example, you can send an email or save a record in an external database.

exports.onExecutePostLogin = async (event, api) => {
  const { riskAssessment } = event.authentication || {};
  const riskIsLow = riskAssessment && riskAssessment.confidence === 'low';

  if (riskIsLow) {
    // log(externalDatabase, riskAssessment);
  }
}

Use the assessments object to access the details for individual assessments, including the code property:

exports.onExecutePostLogin = async (event, api) => {
  const { riskAssessment } = event.authentication || {};
  const { ImpossibleTravel } = riskAssessment && riskAssessment.assessments;

  if (ImpossibleTravel.code === 'impossible_travel_from_last_login') {
    // ...
  }
}

Use the assessments object to access the details for individual assessments, and then use the confidence property, the code property, or both.

To learn more about custom confidence scoring, read Custom confidence scoring.

Use the assessments object to access the details for individual assessments, including the code property.

Block the login transaction from completing by returning the callback function with an UnauthorizedError object as the error parameter. The UnauthorizedError object always sets error to unauthorized, but you can customize the error_message:

exports.onExecutePostLogin = async (event, api) => {
  const { riskAssessment } = event.authentication || {};
  const { ImpossibleTravel } = riskAssessment && riskAssessment.assessments;

  if (ImpossibleTravel.code === 'impossible_travel_from_last_login') {
    return api.access.deny('Login blocked due to impossible travel detected.')
  }
}

This redirects the user back to the application's callback URL with the error and error_message parameters included.

Auth0 automatically assigns a low confidence score if there is any sort of failure performing the risk assessment.

To mitigate this scenario, use the assessments object to inspect the code property for each individual assessment and check if the value is set to assessment_not_available.

• Enable Adaptive MFA
• Adaptive MFA Log Events
• Actions Triggers: post-login - API Object
• Actions Triggers: post-login - Event Object