Access prediction service receiving authentication journey characteristics for supervised learning
A process, system and medium for building a training set and performing supervised training of a Machine Learning (ML) model that determines a risk score used to decide whether to impose stepped up authentication during an authentication journey are described. The process includes selecting examples of completed authentication journeys, including failed and successful authentication outcomes after step-up requirements during the example journeys. The process includes pairing ground truth outcomes from the example journeys with authentication request features initially available prior to imposition of the step-up requirements to produce request feature-outcome pairs. The process includes using at least the request feature-outcome pairs to perform the supervised training of the ML model to determine a risk score that can be used to decide whether to impose the stepped up authentication during an authentication journey. The system and medium are configured to execute the process.
1. A method of building a training set and performing supervised training of a Machine Learning (ML) model that determines a risk score used to decide whether to impose stepped up authentication during an authentication journey, the method comprising:
selecting example completed authentication journeys, including failed and successful authentication outcomes after step-up requirements during the example completed authentication journeys;
pairing ground truth outcomes from the example completed authentication journeys with authentication request features initially available prior to imposition of the step-up requirements to produce request feature-outcome pairs, the authentication request features including 5 or more of:
geographical country from which an authentication request originated,
geographical city from which the authentication request originated,
device type from which the authentication request originated,
day of week on which the authentication request was received,
part of day in which the authentication request was received,
operating system (OS) type from which the authentication request was sent, or
user agent from which the authentication request was sent; and
using at least the request feature-outcome pairs to perform the supervised training of the ML model to produce a trained ML model to determine a risk score that can be used to decide whether to impose the stepped up authentication during an authentication journey.
2. The method of claim 1 , further including:
receiving fresh authentication request features of the authentication journey;
processing, with the trained ML model, the fresh authentication request features; and
providing the risk score determined by the trained ML model.
3. The method of claim 2 , further including:
if the risk score exceeds a threshold, providing an explanation of the risk score, wherein the explanation comprises a feature.
4. The method of claim 1 , wherein the supervised training of the ML model uses journey features as training features.
5. The method of claim 4 , wherein the journey features include type of step-up.
6. The method of claim 5 , wherein the type of step-up comprises at least one of multi-factor authentication, CAPTCHA, or biometrics.
7. A non-transitory computer-readable medium configured with instructions to build a training set and perform supervised training of a Machine Learning (ML) model that determines a risk score used to decide whether to impose stepped up authentication during an authentication journey, the instructions, when executed on a processor, implement a method comprising:
selecting example completed authentication journeys, including failed and successful authentication outcomes after step-up requirements during the example completed authentication journeys;
pairing ground truth outcomes from the example completed authentication journeys with authentication request features initially available prior to imposition of the step-up requirements to produce request feature-outcome pairs, the authentication request features including 5 or more of:
geographical country from which an authentication request originated,
geographical city from which the authentication request originated,
device type from which the authentication request originated,
day of week on which the authentication request was received,
part of day in which the authentication request was received,
operating system (OS) type from which the authentication request was sent, or
user agent from which the authentication request was sent; and
using at least the request feature-outcome pairs to perform the supervised training of the ML model to produce a trained ML model to determine a risk score that can be used to decide whether to impose the stepped up authentication during an authentication journey.
8. The non-transitory computer-readable medium of claim 7 , the method further including:
receiving fresh authentication request features of the authentication journey;
processing, with the trained ML model, the fresh authentication request features; and
providing the risk score determined by the trained ML model.
9. The non-transitory computer-readable medium of claim 8 , the method further including:
if the risk score exceeds a threshold, providing an explanation of the risk score, wherein the explanation comprises a feature.
10. The non-transitory computer-readable medium of claim 7 , wherein the supervised training of the ML model uses journey features as training features.
11. The non-transitory computer-readable medium of claim 10 , wherein the journey features include type of step-up.
12. The non-transitory computer-readable medium of claim 11 , wherein the type of step-up comprises at least one of multi-factor authentication, CAPTCHA, or biometrics.
13. A system including one or more processors coupled to memory, the memory loaded with computer instructions to build a training set and perform supervised training of a Machine Learning (ML) model that determines a risk score used to decide whether to impose stepped up authentication during an authentication journey, the computer instructions, when executed on the one or more processors, implement actions comprising:
selecting example completed authentication journeys, including failed and successful authentication outcomes after step-up requirements during the example completed authentication journeys;
pairing ground truth outcomes from the example completed authentication journeys with authentication request features initially available prior to imposition of the step-up requirements to produce request feature-outcome pairs, the authentication request features including 5 or more of:
geographical country from which an authentication request originated,
geographical city from which the authentication request originated,
device type from which the authentication request originated,
day of week on which the authentication request was received,
part of day in which the authentication request was received,
operating system (OS) type from which the authentication request was sent, or
user agent from which the authentication request was sent; and
using at least the request feature-outcome pairs to perform the supervised training of the ML model to produce a trained ML model to determine a risk score that can be used to decide whether to impose the stepped up authentication during an authentication journey.
14. The system of claim 13 , wherein the memory further includes computer instructions that, when executed on the one or more processors, implement actions comprising:
receiving fresh authentication request features of the authentication journey;
processing, with the trained ML model, the fresh authentication request features; and
providing the risk score determined by the trained ML model.
15. The system of claim 14 , wherein the memory further includes computer instructions that, when executed on the one or more processors, implement actions comprising:
if the risk score exceeds a threshold, providing an explanation of the risk score, wherein the explanation comprises a feature.
16. The system of claim 13 , wherein the supervised training of the ML model uses journey features as training features.
17. The system of claim 16 , wherein the journey features include type of step-up.
18. The system of claim 17 , wherein the type of step-up comprises at least one of multi-factor authentication, CAPTCHA, or biometrics.