US-12619695-B2 - Systems and methods for AI assisted biometric authentication

US12619695B2US 12619695 B2US12619695 B2US 12619695B2US-12619695-B2

Abstract

In one embodiment, a method includes receiving biometrics of a user to access the electronic device. The method may access one or more contextual parameters affecting a state of the user. The method may verify, using a trained machine-learning model, the biometrics of the user based on the one or more contextual parameters affecting the state of the user. The method may provide access to the electronic device in response to successful verification.

Inventors

Alexi Georgiev Jordanov

Assignees

SAMSUNG ELECTRONICS COMPANY, LTD.

Dates

Publication Date: 20260505
Application Date: 20230315

Claims (20)

1 . A method, performed by an electronic device, comprising: receiving biometrics of a user to access the electronic device; accessing one or more contextual parameters affecting a state of the user, wherein the one or more contextual parameters comprise at least one or more external contextual parameters relating to an external environment surrounding the user at a time when the biometrics of the user are received; verifying, using a trained machine-learning model, the biometrics of the user by providing as inputs, to the trained machine-learning model, the biometrics of the user and the at least one or more external contextual parameters relating to the external environment surrounding the user, wherein the trained machine-learning model is trained, for a pre-determined number of training iterations, based on a plurality of user biometrics captured at a plurality of time instances, contextual parameters affecting a user state at each of the plurality of time instances, and ground truth data representing verified authentication results produced for the plurality of user biometrics by one or more of a first conventional technique or a second conventional technique of user authentication, and wherein, at each training iteration, the trained machine-learning model is configured to generate an output based on captured user biometrics at a particular time instance and a set of contextual parameters affecting the user state at the particular time instance; and providing access to the electronic device in response to successful verification.
2 . The method of claim 1 , wherein prior to verifying the biometrics of the user using the trained machine-learning model: verifying the biometrics of the user using the first conventional technique of user authentication; and determining that verification using the first conventional technique is unsuccessful, wherein the verification using the trained machine-learning model is performed in response to determining that the verification using the first conventional technique is unsuccessful.
3 . The method of claim 2 , wherein the first conventional technique of user authentication comprises matching the biometrics with stored biometrics in a data store.
4 . The method of claim 2 , wherein: verifying the biometrics of the user using the first conventional technique comprises: determining whether a match between the biometrics of the user and previously stored biometrics is within a certain threshold; and determining whether a time of last successful verification is within a specific pre-configured time range; and determining that the verification using the first conventional technique is unsuccessful comprises: determining that the match between the biometrics of the user and the previously stored biometrics is beyond the certain threshold; and determining that the time of last successful verification is outside of the specific pre-configured time range.
5 . The method of claim 2 , further comprising: determining that the verification using the trained machine-learning model is unsuccessful; in response to determining that the verification using the trained machine-learning model is unsuccessful, verifying the biometrics of the user using the second conventional technique of user authentication; determining that the verification using the second conventional technique is successful; and re-training the trained machine-learning model based on the biometrics of the user, the one or more contextual parameters, and a label indicating that the biometrics of the user are valid.
6 . The method of claim 5 , wherein the second conventional technique of user authentication comprises requesting the user to manually provide authentication credentials.
7 . The method of claim 5 , wherein re-training the trained machine-learning model comprises updating one or more components of the trained machine-learning model.
8 . The method of claim 2 , further comprising: determining that the verification using the trained machine-learning model is unsuccessful; in response to determining that the verification using the trained machine-learning model is unsuccessful, verifying the biometrics of the user using the second conventional technique of user authentication; determining that the verification using the second conventional technique is unsuccessful; and restricting access to the electronic device in response to unsuccessful verifications using the first conventional technique, the trained machine-learning model, and the second conventional technique.
9 . The method of claim 1 , where the biometrics comprise one or more of fingerprints, retina, face, or voice of the user.
10 . The method of claim 1 , where the state of the user comprises a skin condition of the user.
11 . The method of claim 1 , where the one or more contextual parameters further comprise: internal contextual parameters relating to internal mood or behavior of the user.
12 . The method of claim 1 , wherein the one or more external contextual parameters comprise one or more of geolocation, time of day, season, weather conditions, outside temperature, or humidity.
13 . The method of claim 11 , wherein the internal contextual parameters comprise one or more of user behavior, user mood, user pulse rate, user heart rate, or user expressions.
14 . The method of claim 1 , further comprising training the machine-learning model, wherein training the machine-learning model comprises: receiving, at the plurality of time instances, the plurality of user biometrics; and for each particular time instance of the plurality of time instances, training the machine-learning model by: verifying, using one or more of the first conventional technique or the second conventional technique of user authentication, the captured user biometrics at the particular time instance; determining that the captured user biometrics at the particular time instance are valid based on verifying the biometrics using one or more of the first conventional technique or the second conventional technique; accessing the set of contextual parameters affecting the user state at the particular time instance; and providing the captured user biometrics at the particular time instance and the set of contextual parameters affecting the user state at the particular time instance as inputs for training the machine-learning model for user authentication.
15 . The method of claim 14 , further comprising: enabling the user to select one or more contextual parameters from the set of contextual parameters and weightage for each of the one or more contextual parameters for training the machine-learning model, wherein the machine-learning model is trained based on user selected contextual parameters and associated weightage.
16 . The method of claim 14 , further comprising: enabling the user to enable or disable training of the machine-learning at one or more time instances.
17 . The method of claim 14 , further comprising: enabling the user to choose whether to use trained machine-learning model for user authentication at one or more time instances.
18 . The method of claim 14 , wherein the determination that the captured user biometrics at the particular time instance are valid according to the first conventional technique or the second conventional technique is used as ground truth for training the machine-learning model.
19 . An electronic device comprising: one or more non-transitory computer-readable storage media including instructions; and one or more processors coupled to the storage media, the one or more processors configured to execute the instructions to: receive biometrics of a user to access the electronic device; access one or more contextual parameters affecting a state of the user, wherein the one or more contextual parameters comprise at least one or more external contextual parameters relating to an external environment surrounding the user at a time when the biometrics of the user are received; verify, using a trained machine-learning model, the biometrics of the user by providing as inputs, to the trained machine-learning model, the biometrics of the user and the at least one or more external contextual parameters relating to the external environment surrounding the user, wherein the trained machine-learning model is trained, for a pre-determined number of training iterations, based on a plurality of user biometrics captured at a plurality of time instances, contextual parameters affecting a user state at each of the plurality of time instances, and ground truth data representing verified authentication results produced for the plurality of user biometrics by one or more of a first conventional technique or a second conventional technique of user authentication, and wherein, at each training iteration, the trained machine-learning model is configured to generate an output based on captured user biometrics at a particular time instance and a set of contextual parameters affecting the user state at the particular time instance; and provide access to the electronic device in response to successful verification.
20 . A non-transitory computer-readable medium comprising instructions that, when executed by one or more processors of an electronic device, cause the one or more processors to: receive biometrics of a user to access the electronic device; access one or more contextual parameters affecting a state of the user, wherein the one or more contextual parameters comprise at least one or more external contextual parameters relating to an external environment surrounding the user at a time when the biometrics of the user are received; verify, using a trained machine-learning model, the biometrics of the user by providing as inputs, to the trained machine-learning model, the biometrics of the user and the at least one or more external contextual parameters relating to the external environment surrounding the user, wherein the trained machine-learning model is trained, for a pre-determined number of training iterations, based on a plurality of user biometrics captured at a plurality of time instances, contextual parameters affecting a user state at each of the plurality of time instances, and ground truth data representing verified authentication results produced for the plurality of user biometrics by one or more of a first conventional technique or a second conventional technique of user authentication, and wherein, at each training iteration, the trained machine-learning model is configured to generate an output based on captured user biometrics at a particular time instance and a set of contextual parameters affecting the user state at the particular time instance; and provide access to the electronic device in response to successful verification.

Description

TECHNICAL FIELD This disclosure relates generally to database and file management within network environments, and in particular relates to verifying or authenticating biometrics of a user using a machine-learning model. BACKGROUND Nowadays people use their electronic devices for a variety of purposes. For instance, a user may use an electronic device to check their messages, make phone calls, check social feeds, make social interactions, capture images, record videos, etc. Such electronic devices may include, for example and without limitation, smartphones, tablets, computers, smartwatches, and so forth. Generally, a user of an electronic device may be requested to provide their biometrics to access the device. For instance, the user may be asked to provide their fingerprints by touching at a specific location on the device and the device may grant access (e.g., unlock device) if the device is successfully able to authenticate or validate the fingerprints. Sometimes, the device is not able to authenticate the biometrics of the user even when the user is in fact a genuine user. For instance, human biometrics are subject to change due to different natural factors. For example, skin may dry up and therefore changing the fingerprint characteristics. As a result, applying the fingerprint to unlock the mobile device may fail. Also, human biometrics may be applied in a way that resulting biometric vector is different from a recorded biometric vector (e.g., changing the angle of input biometrics) and as a result, the authentication attempt may incorrectly fail. As such, there is a need for a technique for biometric authentication of a user under such failed authentication attempts and/or circumstances. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 illustrates an example scenario of biometric authentication of a user at two different time instances. FIG. 2 illustrates an example electronic device. FIG. 3 illustrates an overall environment for biometric authentication of a user. FIG. 4 illustrates an example environment for training a machine-learning model for biometric authentication. FIG. 5 illustrates an example signal or interaction flow diagram depicting example interactions between various components of an electronic device for biometric authentication. FIG. 6 illustrates a flow diagram of an example method for biometric authentication of a user using a trained machine-learning model. FIG. 7 illustrates an example computer system. FIG. 8 illustrates a diagram of an example artificial intelligence (AI) architecture. DESCRIPTION OF EXAMPLE EMBODIMENTS Biometric authentication is an essential way to get access to an electronic device, such as a smartphone, a tablet, a computer, a smartwatch, an artificial-reality system, etc. Some of the biometric authentication methods may include, for example and without limitation, fingerprint matching and sensing, credentials check (e.g., username, pin, password, etc.), retina scanning, face tracking, etc. Sometimes a user may have a hard time validating or authenticating their biometrics on their electronic device. For instance, the user may be able to validate their biometrics (e.g., fingerprints) at one time instance, but they may not be able to validate at a second time instance. Failure in such biometric authentication may be due to contextual parameters affecting or impacting a state of the user. These contextual parameters affecting the state of the user may include one or more of external contextual parameters relating to external environment surrounding the user or internal contextual parameters relating to internal mood or behavior of the user. The external contextual parameters may include, for example and without limitation, geolocation, time of day, season, weather conditions, outside temperature, humidity, etc. The internal contextual parameters may include, for example and without limitation, user behavior, user mood, user pulse rate, user heart rate, or user expressions. These external and/or internal contextual parameters may lead to failure in biometric authentication. As an example and not by way of limitation, different weather conditions, time of day, or season of the year may be causing skin dryness due to which a fingerprint sensor on the device is not able to correctly recognize the fingerprints of the user. One such example scenario is depicted in FIG. 1, as discussed below. FIG. 1 illustrates an example scenario 100 of biometric authentication of a user at two different time instances 110 and 120. At the first time instance 110, it is raining and the user may be inside their home 112. When the user tries to unlock their device 114 by providing their fingerprints 116, the device 114 is able to successfully authenticate the user. This may be because the fingerprints 116 of the user might be all neat and clean as they are sitting in their home watching television, even though it is raining outside. At the second time instance 120, the user may be now out in