Generative artificial intelligence for validation of a human user

Patent metadata

Field	Value
Publication number	US-12572639-B2
Application number	US-202418635080-A
Country	US
Kind code	B2
Filing date	Apr 15, 2024
Priority date	Apr 15, 2024
Publication date	Mar 10, 2026
Grant date	Mar 10, 2026

Abstract

There is provided a computer implemented method of validation of a human user, comprising: feeding a description of a first object into a GenAI model to generate at least one image depicting a plurality of instances of the first object and a plurality of second objects different than the first object, each instance representing a unique variation of the first object, via a user interface presented on a display of a client terminal: presenting the at least one image, presenting instructions for a user to identify common instances of objects having unique variations that are depicted in the at least one image, receiving the user indication, and validating that the user is a human when the common instances of objects identified by the user matches the description of the first object fed into the GenAI model.

First claim

What is claimed is:

1 . A computer implemented method of validation of a human user, comprising:
feeding a description of a first object into a GenAI model to generate at least one image depicting a plurality of instances of the first object and a plurality of second objects different than the first object, each instance representing a unique variation of the first object; via a user interface presented on a display of a client terminal:
presenting the at least one image;
presenting instructions for a user to identify common instances of objects having unique variations that are depicted in the at least one image;
receiving the user indication; and
validating that the user is a human when the common instances of objects identified by the user matches the description of the first object fed into the GenAI model; wherein the at least one image comprises a plurality of images, and a number of the plurality of images and the instructions for the user for validating that the user is human, are selected according to a statistical prediction of the GenAI model erroneously generating an image of the number of the plurality of images that does not match the description of the first object, wherein the validating that the user is human is performed when a target less than the number of images is selected by the user for compensating for the statistical prediction of the GenAI model erroneously generating the image.

2 . The computer implemented method of claim 1 , wherein the description of the first object fed into the GenAI model is non-presented on the display and/or is non-accessible to the user.

3 . The computer implemented method of claim 1 , wherein the GenAI model generates a plurality of first images each including a respective instance of the first object and a plurality of second images each excluding the first object, and the presented instructions are to identify a set of images from the plurality of first images and the plurality of second images that have a common object.

4 . The computer implemented method of claim 1 , wherein the GenAI model generates a plurality of first images each including a respective instance of the first object and a plurality of second images each excluding the first object, and the presented instructions are to identify a set of images from the plurality of first images and the plurality of second images that do not have an object in common.

5 . The computer implemented method of claim 1 , wherein the unique variations are selected from: a unique angle, a unique size, a unique shape, a unique additional structural element, depicted as being made from a unique material, a certain type of the first object located on and/or within another type of the first object, a unique texture, a unique color, and depicted as made from a third object that is visually similar to a type of the first object.

6 . The computer implemented method of claim 1 , wherein the description including the first object is randomly generated for creating a unique image.

7 . The computer implemented method of claim 1 , wherein the description indicates a scenario including the first object that is physically impossible in real life.

8 . The computer implemented method of claim 1 , wherein the description is of a representation of the first object that is different than the first object itself.

9 . The computer implemented method of claim 8 , wherein the description is selected from a group comprising: an embroidery texture of the first object, a comic illustration of the first object, a non-realistic painted version of the first object, and a representation of the first object on a surface of a third object.

10 . The computer implemented method of claim 1 , wherein the description is of a texture of the first object that is different than a real life texture of the first object.

11 . The computer implemented method of claim 1 , wherein the at least one image further depicts a plurality of second objects different than the first object.

12 . The computer implemented method of claim 1 , wherein the at least one image comprises a first image generated using a first random seed, and further comprising:
generating a second image by feeding the description into the GenAI using a second random seed, and generating a plurality of third images depicting a plurality of second objects different than the first object using the GenAI model; wherein presenting the image comprises presenting the first image, the second image, and the plurality of third images, wherein the presented instructions are for matching images with a common element, wherein validating comprises validating when the first image and the second image are indicated by the user.

13 . The computer implemented method of claim 1 , further comprising iterating the feeding, the presenting the at least one image, the presenting instructions, and the receiving instructions, for validating that the user is human when a number of iterations in which the common instances of objects identified by the user correctly matches the description, is selected according to a statistical prediction of the GenAI model erroneously generating the at least one image when the at least one image does not match the description of the first object.

14 . The computer implemented method of claim 1 , wherein at least one of the number of the plurality of images is intentionally created using a different description of the first object.

15 . A computer implemented method of validation of a human user, comprising:
feeding a description of an object into a generative artificial intelligence (GenAI) model to generate at least one image with a representation of the object that is different than the object itself; via a user interface presented on a display of a client terminal:
presenting the at least one image;
presenting instructions for a user to indicate the object,
wherein the description of the object fed into the GenAI model is non-presented on the display and/or is non-accessible to the user;
receiving the user indication; and
validating that the user is a human when the user indication matches the description fed into the GenAI model; wherein the at least one image comprises a plurality of images, and a number of the plurality of images and the instructions for the user for validating that the user is human, are selected according to a statistical prediction of the GenAI model erroneously generating an image of the number of the plurality of images that does not match the description of the first object, wherein the validating that the user is human is performed when a target less than the number of images is selected by the user for compensating for the statistical prediction of the GenAI model erroneously generating the image.

16 . The computer implemented method of claim 15 , wherein the description is selected from a group comprising: an embroidery texture of the object, a comic illustration of the object, a non-realistic painted version of the object, and a representation of the object on a surface of a second object.

17 . The computer implemented method of claim 15 , wherein the description is of a texture of the object that is different than a real life texture of the object.

18 . A computer implemented method of validation of a human user, comprising:
accessing at least one personal parameter of a user; feeding a text directive including the at least one personal parameter into a generative artificial intelligence (GenAI) model to generate at least one image visually depi

Assignees

• Nec Corp America

Inventors

• Lev Tsvi
• Hoch Yaacov

Classifications

• G06F2221/2133

  Verifying human interaction, e.g., Captcha · CPC title

• G06F21/36 Primary

  by graphic or iconic representation · CPC title