Granted Patent B2

US 12,335,108 · App. 18/219,352 · Granted Jun 17, 2025

Inventors: Padmapriya Mohankumar (Chennai, IN); Vishal Kumar Singh (Howrah, IN); Ashraf Kamal (New Delhi, IN)

Assignee: PAYPAL, INC.

H04L41/16

Abstract

Claims (52)

1. A computer-implemented method comprising:

providing an embedded service including a first embedded service and a second embedded service in an application;

training a monitoring model via federated learning based on data derived locally from the application;

monitoring, via the trained monitoring model, a plurality of interactions with the embedded service;

determining, by the trained monitoring model, that at least one of the plurality of interactions comprises an anomalous interaction; and

in response to the determination, restricting further usage of the application,

wherein the federated learning comprises training a local model based on the data derived locally from the application, and the monitoring model is trained based on results from training the local model, and

wherein training the monitoring model via federated learning comprises:

training a first iteration of the monitoring model with a first set of historical interactions;

retrieving a first parameters from the first trained iteration;

training a second iteration of the monitoring model with a second set of historical interactions;

retrieving a second parameters from the second trained iteration; and

determining parameters for the trained monitoring model as an average of the first parameters and the second parameters.

2. The method of claim 1 , wherein:

training the monitoring model via federated learning comprises:

retrieving the first set of historical interactions with the first embedded service; and retrieving the second set of historical interactions with the second embedded service.

3. The method of claim 1 , wherein:

the first embedded service is provided to a first user device,

the second embedded service is provided to a second user device, and

the first user device and the second user device are separate devices operating a same application.

4. The method of claim 1 , wherein the first iteration and the second iteration are each trained independently of the other of the first iteration and the second iteration.

5. The method of claim 1 , wherein:

the application comprises a first application,

the method further comprises providing the first embedded service in the first application and the second embedded service in a second application, and

training the monitoring model via federated learning comprises:

retrieving the first set of historical interactions with the first embedded service in the first application; and

retrieving the second set of historical interactions with the second embedded service in the second application.

6. The method of claim 5 , wherein the first iteration and the second iteration are each trained independently of the other of the first iteration and the second iteration.

7. The method of claim 1 , wherein the embedded service comprises a contract review module configured to:

receive a document having a plurality of terms;

process the document to determine a critical term of the plurality of terms; and

present, via the application, an action in response to the critical term.

8. The method of claim 1 , further comprising:

processing the data derived locally from the application to determine a pattern of interactions with the embedded service; and

presenting, on the application, a recommend action in the embedded service based on the determined pattern.

9. The method of claim 1 , wherein restricting further usage of the application comprises restricting further usage by a particular device.

10. A computer-implemented method comprising:

providing, by a server, an application to a first device and to a second device, the application having an embedded service;

storing, on the first device, a first dataset indicative of interactions with the embedded service by the first device;

storing, on the second device, a second dataset indicative of interactions with the embedded service by the second device;

training, locally on the first device, a first portion of a machine learning model using the first dataset;

training, locally on the second device, a second portion of the machine learning model using the second dataset;

retrieving, by the server from the first device and the second device, respective parameters associated with the trained first portion and the trained second portion, respectively;

adjusting the machine learning model based on the retrieved parameters;

determining, via the adjusted machine learning model, that an interaction with the application comprises an anomaly; and

in response to the determined anomaly, restricting access to the application for a device associated with the determined anomaly,

wherein adjusting the machine learning model based on the retrieved parameters comprises:

determining a shared set of parameters as an average of the parameters associated with the trained first portion and the parameters associated with the trained second portion; and

applying the shared set of parameters to the machine learning model.

11. The method of claim 10 , wherein the first portion and the second portion are trained independently of the other of the first portion and the second portion.

12. The method of claim 10 , wherein the parameters received from the first device are different from the parameters received from the second device.

13. The method of claim 10 , wherein the first portion of the machine learning model is identical to the second portion of the machine learning model before training by the first device and the second device.

Assignments (1)

Continuity (1)

References Cited (22)