Systems and methods of generating risk scores and predictive fraud modeling
One or more implementations include methods, systems, and/or devices to help protect consumers from fraudulent activity using compromised PII. For example, systems and methods can be implemented that enable the detection and prevention of consumer-focused identity theft, generates a risk score and is powered by a predictive model using machine learning techniques and tools and presents information and recommended action a user can take in reports.
1. A system comprising:
one or more data stores configured to store:
computer-executable instructions;
exposed personal identifiable information (PII) collected from one or more internet-accessible sources, wherein the exposed PII corresponds to a plurality of users;
training PII associated with confirmed fraud cases and confirmed non-fraud cases; and
first user PII associated with a first user;
a network interface configured to communicate with a plurality of network devices; and
one or more physical computer processors in communication with the one or more data stores, wherein the computer-executable instructions, when executed, configure the one or more physical computer processors to:
access, by the network interface and from the one or more data stores, the exposed PII, the training PII, and the first user PII;
generate a first predictive model by training a first machine learning algorithm, wherein the training comprises inputting the exposed PII corresponding to the plurality of users and the training PII into the first machine learning algorithm, wherein the training further includes comparing the training PII with a corresponding first subset of the exposed PII, wherein the first predictive model is configured to determine a likelihood of future fraudulent activities based on identifying data elements associated with confirmed fraud cases and confirmed non-fraud cases;
provide, into the first predictive model, a first input comprising the first user PII and a corresponding second subset of the exposed PII, the second subset including PII associated with the first user;
receive, from the first predictive model, a first output that includes a likelihood of future fraudulent activities associated with the first input;
generate a first risk score and first recommendations for the first user based on the first output; and
based at least in part on the first risk score, generate and transmit, to a first user device associated with the first user, display instructions configured to present an interactive user interface comprising a risk report, wherein the risk report comprises the first risk score and first recommendations.
2. The system of claim 1 , wherein training the first machine learning algorithm further comprises:
determining a first set of variables by comparing the exposed PII to the training PII;
applying variable or feature importance to the first set of variables to generate a feature importance score for each variable of the first set of variables;
determining a second set of variables based at least in part on application of the variable or feature importance, wherein the second set of variables is a subset of the first set of variables; and
generating the first predictive model based on the second set of variables.
3. The system of claim 1 , wherein the first recommendations include one or more of: closing an account, changing a password, changing a username, changing or removing PII associated with an account, changing multi-factor authentication settings, and changing general settings on an account.
4. The system of claim 1 , wherein the first recommendations include recommended actions the first user can take to improve the first risk score or reduce a likelihood of future fraudulent activity with respect to the first user PII.
5. The system of claim 4 , wherein the one or more physical computer processors are further configured to:
monitor status of the first recommendations; and
based on the monitoring, determine that a second recommendation of the first recommendations has been completed.
6. The system of claim 5 , wherein the one or more physical computer processors are further configured to:
in response to the determination that the second recommendation has been completed, generate an updated risk score and updated recommendations for the first user.
7. The system of claim 1 , wherein the risk report further comprises one or more of: a subpopulation risk score, a user password analysis, a percent contribution analysis, a list of riskiest breached sites, and a web exposure analysis.
8. The system of claim 1 , wherein confirmed fraud cases and confirmed non-fraud cases are determined by data collected from one or more of: one or more proprietary databases, one or more users, and one or more third party services or companies.
9. The system of claim 2 , wherein the first predictive model comprises an equation comprising weighted coefficients and variables, wherein the variables are based at least in part on the second set of variables.
10. The system of claim 2 , wherein the comparing of the exposed PII to the training PII comprises matching PII from the exposed PII collected from one or more internet-accessible sources with PII from the training PII associated with confirmed fraud cases and confirmed non-fraud cases.
11. The system of claim 2 , wherein determining the second set of variables includes determining all variables of the first set of variables that comprise a feature importance score that meets a preconfigured threshold score that indicates a threshold predictive value for the first predictive model.
12. The system of claim 2 , wherein the second set of variables includes fewer variables than the first set of variables.
13. A computer-implemented method comprising:
accessing, by a network interface and from one or more data stores:
exposed personal identifiable information (PII) collected from one or more internet-accessible sources, wherein the exposed PII corresponds to a plurality of users;
training PII associated with confirmed fraud cases and confirmed non-fraud cases; and
first user PII associated with a first user;
generating a first predictive model by training a first machine learning algorithm, wherein the training comprises inputting the exposed PII corresponding to the plurality of users and the training PII into the first machine learning algorithm, wherein the training further includes comparing the training PII with a corresponding first subset of the exposed PII, wherein the first predictive model is configured to determine a likelihood of future fraudulent activities based on identifying data elements associated with confirmed fraud cases;
providing, into the first predictive model, a first input comprising the first user PII and a corresponding second subset of the exposed PII, the second subset including PII associated with the first user;
receiving, from the first predictive model, a first output that includes a likelihood of future fraudulent activities associated with the first input;
generating a first risk score and first recommendations for the first user based on the first output; and
based at least in part on the first risk score, generating and transmitting, to a first user device associated with the first user, display instructions configured to present an interactive user interface comprising a risk report, wherein the risk report comprises the first risk score and first recommendations.
14. The method of claim 13 , wherein training the first machine learning algorithm further comprises:
determining a first set of variables by comparing the exposed PII to the training PII;
applying variable or feature importance to the first set of variables to generate a feature importance score for each variable of the first set of variables;
determining a second set of variables based at least in part on application of the variable or feature importance, wherein the second set of variables is a subset of the first set of variables; and
generating the first predictive model based on the second set of variables.
15. The method of claim 13 , wherein the first recommendations include recommended actions the first user can take to improve the first risk score or reduce a likelihood of future fraudulent activity with respect to the first user PII.
16. The method of claim 14 , wherein the comparing of the exposed PII to the training PII comprises matching PII from the exposed PII collected from one or more internet-accessible sources with PII from the training PII associated with confirmed fraud cases and confirmed non-fraud cases.
17. The method of claim 14 , wherein determining the second set of variables includes determining all variables of the first set of variables that comprise a feature importance score that meets a preconfigured threshold score that indicates a threshold predictive value for the first predictive model.
18. A non-transitory computer storage medium storing computer-executable instructions that, when executed by a processor, cause the processor to at least:
access, by a network interface and from one or more data stores:
exposed personal identifiable information (PII) collected from one or more internet-accessible sources, wherein the exposed PII correspond to a plurality of users;
training PII associated with confirmed fraud cases and confirmed non-fraud cases; and
first user PII associated with a first user;
generate a first predictive model by training a first machine learning algorithm, wherein the training comprises inputting the exposed PII corresponding to the plurality of users and the training PII into the first machine learning algorithm, wherein the training further includes comparing the training PII with a corresponding first subset of the exposed PII, wherein the first predictive model is configured to determine a likelihood of future fraudulent activities;
provide, into the first predictive model, a first input comprising the first user PII and a corresponding second subset of the exposed PII, the second subset including PII associated with the first user;
receive, from the first predictive model, a first output that includes a likelihood of future fraudulent activities associated with the first input;
generate a first risk score and first recommendations for the first user based on the first output; and
based at least in part on the first risk score, generate and transmit, to a first user device associated with the first user, display instructions configured to present an interactive user interface comprising a risk report, wherein the risk report comprises the first risk score and first recommendations.
19. The non-transitory computer storage medium of claim 18 , wherein training the first machine learning algorithm further comprises:
determining a first set of variables by comparing the exposed PII to the training PII;
applying variable or feature importance to the first set of variables to generate a feature importance score for each variable of the first set of variables;
determining a second set of variables based at least in part on application of the variable or feature importance, wherein the second set of variables is a subset of the first set of variables; and
generating the first predictive model based on the second set of variables.
20. The system of claim 1 , wherein the risk report further comprises one or more of: exposure details that include quantities of exposed first user information, password analytics that include user password analysis, source analytics that include information on the one or more internet-accessible sources where the exposed PII was collected from, and PII exposures that include quantities of exposed first user PII.