US 12,462,052 B2
Using contextual security challenges to prevent bot attacks
Yu Wang, Toronto (CA); Karan Singh Chhina, Forest Hills, NY (US); Ian Hassard, Waterloo (CA); Ankur Priyadarshi Chauhan, Vallauris (FR); Ria Prakash, Toronto (CA); Matthew James Pereira, Halifax (CA); Alec Friedman, East Northport, NY (US); Joseph George Sokolon, Toronto (CA); and Shaun Michael Tarves, Marlborough, MA (US)
Filed by OKTA, INC., San Francisco, CA (US)
Filed on Oct. 18, 2023, as Appl. No. 18/489,794.
Prior Publication US 2025/0131110 A1, Apr. 24, 2025
Int. Cl. G06F 21/60 (2013.01); G06F 21/36 (2013.01)
CPC G06F 21/604 (2013.01) [G06F 21/36 (2013.01); G06F 2221/2103 (2013.01); G06F 2221/2133 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method at a fraud detection service that comprises at least one processor, comprising:
outputting, to a first machine learning model, a first set of prompts, the first set of prompts corresponding to a plurality of image descriptions, and each prompt of the first set of prompts corresponding to a request of the first machine learning model to generate a respective image description of the plurality of image descriptions;
obtaining, from the first machine learning model, information indicative of the plurality of image descriptions generated via the first machine learning model in response to the request to generate the plurality of image descriptions indicated via the first set of prompts, each image description of the plurality of image descriptions describing a respective image of a plurality of images, wherein the plurality of image descriptions comprises a first image description set that corresponds to a sequence of events and a second image description set that is unassociated with the sequence of events;
outputting, to a second machine learning model, a second set of prompts different from the first set of prompts, each prompt of the second set of prompts corresponding to a request of the second machine learning model to generate a respective image of the plurality of images based on the respective image description;
obtaining, from the second machine learning model, the plurality of images generated via the second machine learning model in response to the request to generate the plurality of images indicated via the second set of prompts, wherein the plurality of images comprises a first image set that corresponds to the sequence of events based at least in part on the first image description set, and a second image set that is unassociated with the sequence of events based at least in part on the second image description set; and
generating, via the at least one processor of the fraud detection service, a security challenge using the plurality of images, wherein the security challenge comprises a request for a user to identify the sequence of events from among the plurality of images, and wherein identification of the sequence of events is based at least in part on each image of the first image set being contextually relevant to the sequence of events.