US-20260127487-A1 - System and method to dynamically detect anomalies in data exchange operations

Abstract

A system comprises a memory communicatively coupled to at least one processor. The at least one processor is configured to receive a request to perform a data exchange operation at a communication level and an evaluation level and determine multiple configuration parameters based on the data exchange operation and execute one or more machine learning algorithms to determine an operational path to complete the data exchange operation over a path duration based on the configuration parameters, assign a static token to the data exchange operation based on the communication level, assign a dynamic token to the data exchange operation based on the evaluation level, train a communication model to perform one or more sub-operations of the operational path using the static token, the dynamic token, and the configuration parameters, and perform one or more of the sub-operations in accordance with the communication model.

Inventors

• Pratap Dande
• Naga Vamsi Krishna Akkapeddi

Assignees

• BANK OF AMERICA CORPORATION

Dates

Publication Date

20260507

Application Date

20241105

Claims (20)

1. 1 . An apparatus, comprising: a memory operable to store: one or more machine learning algorithms configured to perform one or more operations in accordance with one or more machine learning models; and at least one processor communicatively coupled to the memory and configured to: receive a first request to perform a first data exchange operation at a first communication level and a first evaluation level; determine a first plurality of configuration parameters based on the first data exchange operation, the first plurality of configuration parameters comprising first guidance to perform the first data exchange operation; and execute the one or more machine learning algorithms to: determine a first operational path to complete the first data exchange operation over a first path duration based on the first plurality of configuration parameters, wherein: the first operational path comprises a first plurality of sub-operations to be performed over the first path duration to complete the first data exchange operation; and the first plurality of sub-operations comprising a first sub-operation and a second sub-operation; assign a first static token to the first data exchange operation based on the first communication level, the first static token referencing the first plurality of sub-operations to be performed on the first operational path over the first path duration; assign a first dynamic token to the first data exchange operation based on the first evaluation level, the first dynamic token referencing starting data to be modified by the first plurality of sub-operations on the first operational path over the first path duration; create a first plurality of tolerances for one or more possible changes to the starting data based on the first communication level and the first evaluation level, wherein the first plurality of tolerances comprises: a first tolerance corresponding to a first possible change to the starting data at the first sub-operation; and a second tolerance corresponding to a second possible change to the starting data at the second sub-operation; train a first communication model to perform the first plurality of sub-operations of the first operational path using the first static token, the first dynamic token, the first plurality of tolerances, and the first plurality of configuration parameters; and perform one or more of the first plurality of sub-operations in accordance with the first communication model.
2. 2 . The apparatus of claim 1 , wherein the at least one processor is further configured to: in conjunction with performing one or more of the first plurality of sub-operations in accordance with the first communication model, perform the first sub-operation; in response to performing the first sub-operation, update the first dynamic token to reference a first plurality of changes made to the starting data at the first sub-operation; extract the first plurality of changes made to the starting data at the first sub-operation; determine whether the first plurality of changes is within the first tolerance of the first plurality of tolerances; and in response to determining that the first plurality of changes is within the first tolerance of the first plurality of tolerances, perform the second sub-operation.
3. 3 . The apparatus of claim 1 , wherein the at least one processor is further configured to: in conjunction with performing one or more of the first plurality of sub-operations in accordance with the first communication model, perform the first sub-operation; in response to performing the first sub-operation, update the first dynamic token to reference a first plurality of changes made to the starting data at the first sub-operation; extract the first plurality of changes made to the starting data at the first sub-operation; determine whether the first plurality of changes is within the first tolerance of the first plurality of tolerances; in response to determining that the first plurality of changes is not within the first tolerance of the first plurality of tolerances, determine a communication anomaly associated with the first plurality of changes; determine one or more knowledge base commands configured to fix the communication anomaly; update the first plurality of configuration parameters to account for the one or more knowledge base commands; determine a second operational path to perform the first data exchange operation over a second path duration based on an updated version of the first plurality of configuration parameters; assign a second static token to the first data exchange operation based on the first communication level, the second static token referencing the first plurality of sub-operations to be performed on the second operational path over the second path duration; assign a second dynamic token to the first data exchange operation based on the first evaluation level, the second dynamic token referencing the starting data to be modified by the first plurality of sub-operations on the second operational path over the second path duration; train the first communication model to perform the first plurality of sub-operations of the second operational path using the second static token, the second dynamic token, the first plurality of tolerances, and the updated version of the first plurality of configuration parameters; and perform one or more of the first plurality of sub-operations in accordance with an updated version of the first communication model.
4. 4 . The apparatus of claim 3 , wherein the at least one processor is further configured to: use the updated version of the first communication model to train the one or more machine learning models.
5. 5 . The apparatus of claim 1 , wherein: the first communication level is an enriched communication level that comprises tracking inputs and outputs associated with the first data exchange operation from an origin point to a completion point in the first operational path.
6. 6 . The apparatus of claim 1 , wherein: the first communication level is a granularity communication level that comprises tracking inputs and outputs associated with each of the first plurality of sub-operations at each point in the first operational path.
7. 7 . The apparatus of claim 1 , wherein: the first evaluation level is a request level that causes the first dynamic token to reference the starting data to be modified by all of the first plurality of sub-operations.
8. 8 . The apparatus of claim 1 , wherein: the first evaluation level is a record level that causes the first dynamic token to reference the starting data to be modified by each of the first plurality of sub-operations.
9. 9 . The apparatus of claim 1 , wherein: the first evaluation level is a data-element level that causes the first dynamic token to reference the starting data to be modified by at least one of the first plurality of sub-operations.
10. 10 . The apparatus of claim 1 , wherein the at least one processor is further configured to: receive a second request to perform a second data exchange operation at a second communication level and a second evaluation level; determine a second plurality of configuration parameters based on the second data exchange operation, the second plurality of configuration parameters comprising second guidance to perform the second data exchange operation; and execute the one or more machine learning algorithms to: determine a second operational path to complete the second data exchange operation over a second path duration based on the second plurality of configuration parameters, wherein: the second operational path comprises a second plurality of sub-operations to be performed over the second path duration to complete the second data exchange operation; and the second plurality of sub-operations comprising a third sub-operation and a fourth sub-operation; assign a second static token to the second data exchange operation based on the second communication level, the second static token referencing the second plurality of sub-operations to be performed on the second operational path over the second path duration; assign a second dynamic token to the second data exchange operation based on the second evaluation level, the second dynamic token referencing additional starting data to be modified by the second plurality of sub-operations on the second operational path over the second path duration; create a second plurality of tolerances for one or more additional possible changes to the additional starting data based on the second communication level and the second evaluation level, wherein the second plurality of tolerances comprises: a third tolerance corresponding to a third possible change to the additional starting data at the third sub-operation; and a fourth tolerance corresponding to a fourth possible change to the additional starting data at the fourth sub-operation; train a second communication model to perform the second plurality of sub-operations of the second operational path using the second static token, the second dynamic token, the second plurality of tolerances, and the second plurality of configuration parameters; and perform one or more of the second plurality of sub-operations in accordance with the second communication model.
11. 11 . A method, comprising: receiving a first request to perform a first data exchange operation at a first communication level and a first evaluation level; determining a first plurality of configuration parameters based on the first data exchange operation, the first plurality of configuration parameters comprising first guidance to perform the first data exchange operation; and executing one or more machine learning algorithms to perform one or more operations comprising: determining a first operational path to complete the first data exchange operation over a first path duration based on the first plurality of configuration parameters, wherein: the first operational path comprises a first plurality of sub-operations to be performed over the first path duration to complete the first data exchange operation; and the first plurality of sub-operations comprising a first sub-operation and a second sub-operation; assigning a first static token to the first data exchange operation based on the first communication level, the first static token referencing the first plurality of sub-operations to be performed on the first operational path over the first path duration; assigning a first dynamic token to the first data exchange operation based on the first evaluation level, the first dynamic token referencing starting data to be modified by the first plurality of sub-operations on the first operational path over the first path duration; creating a first plurality of tolerances for one or more possible changes to the starting data based on the first communication level and the first evaluation level, wherein the first plurality of tolerances comprises: a first tolerance corresponding to a first possible change to the starting data at the first sub-operation; and a second tolerance corresponding to a second possible change to the starting data at the second sub-operation; training a first communication model to perform the first plurality of sub-operations of the first operational path using the first static token, the first dynamic token, the first plurality of tolerances, and the first plurality of configuration parameters; and performing one or more of the first plurality of sub-operations in accordance with the first communication model.
12. 12 . The method of claim 11 , further comprising: in conjunction with performing one or more of the first plurality of sub-operations in accordance with the first communication model, performing the first sub-operation; in response to performing the first sub-operation, updating the first dynamic token to reference a first plurality of changes made to the starting data at the first sub-operation; extracting the first plurality of changes made to the starting data at the first sub-operation; determining whether the first plurality of changes is within the first tolerance of the first plurality of tolerances; and in response to determining that the first plurality of changes is within the first tolerance of the first plurality of tolerances, performing the second sub-operation.
13. 13 . The method of claim 11 , further comprising: in conjunction with performing one or more of the first plurality of sub-operations in accordance with the first communication model, performing the first sub-operation; in response to performing the first sub-operation, updating the first dynamic token to reference a first plurality of changes made to the starting data at the first sub-operation; extracting the first plurality of changes made to the starting data at the first sub-operation; determining whether the first plurality of changes is within the first tolerance of the first plurality of tolerances; in response to determining that the first plurality of changes is not within the first tolerance of the first plurality of tolerances, determining a communication anomaly associated with the first plurality of changes; determining one or more knowledge base commands configured to fix the communication anomaly; updating the first plurality of configuration parameters to account for the one or more knowledge base commands; determining a second operational path to perform the first data exchange operation over a second path duration based on an updated version of the first plurality of configuration parameters; assigning a second static token to the first data exchange operation based on the first communication level, the second static token referencing the first plurality of sub-operations to be performed on the second operational path over the second path duration; assigning a second dynamic token to the first data exchange operation based on the first evaluation level, the second dynamic token referencing the starting data to be modified by the first plurality of sub-operations on the second operational path over the second path duration; training the first communication model to perform the first plurality of sub-operations of the second operational path using the second static token, the second dynamic token, the first plurality of tolerances, and the updated version of the first plurality of configuration parameters; and performing one or more of the first plurality of sub-operations in accordance with an updated version of the first communication model.
14. 14 . The method of claim 13 , further comprising: using the updated version of the first communication model to train one or more machine learning models.
15. 15 . The method of claim 11 , wherein: the first communication level is an enriched communication level that comprises tracking inputs and outputs associated with the first data exchange operation from an origin point to a completion point in the first operational path.
16. 16 . A non-transitory computer-readable medium storing instructions that when executed by a processor cause the processor to: receive a first request to perform a first data exchange operation at a first communication level and a first evaluation level; determine a first plurality of configuration parameters based on the first data exchange operation, the first plurality of configuration parameters comprising first guidance to perform the first data exchange operation; and execute one or more machine learning algorithms to: determine a first operational path to complete the first data exchange operation over a first path duration based on the first plurality of configuration parameters, wherein: the first operational path comprises a first plurality of sub-operations to be performed over the first path duration to complete the first data exchange operation; and the first plurality of sub-operations comprising a first sub-operation and a second sub-operation; assign a first static token to the first data exchange operation based on the first communication level, the first static token referencing the first plurality of sub-operations to be performed on the first operational path over the first path duration; assign a first dynamic token to the first data exchange operation based on the first evaluation level, the first dynamic token referencing starting data to be modified by the first plurality of sub-operations on the first operational path over the first path duration; create a first plurality of tolerances for one or more possible changes to the starting data based on the first communication level and the first evaluation level, wherein the first plurality of tolerances comprises: a first tolerance corresponding to a first possible change to the starting data at the first sub-operation; and a second tolerance corresponding to a second possible change to the starting data at the second sub-operation; train a first communication model to perform the first plurality of sub-operations of the first operational path using the first static token, the first dynamic token, the first plurality of tolerances, and the first plurality of configuration parameters; and perform one or more of the first plurality of sub-operations in accordance with the first communication model.
17. 17 . The non-transitory computer-readable medium of claim 16 , wherein, when executed by the processor, the instructions further cause the processor to: in conjunction with performing one or more of the first plurality of sub-operations in accordance with the first communication model, perform the first sub-operation; in response to performing the first sub-operation, update the first dynamic token to reference a first plurality of changes made to the starting data at the first sub-operation; extract the first plurality of changes made to the starting data at the first sub-operation; determine whether the first plurality of changes is within the first tolerance of the first plurality of tolerances; and in response to determining that the first plurality of changes is within the first tolerance of the first plurality of tolerances, perform the second sub-operation.
18. 18 . The non-transitory computer-readable medium of claim 16 , wherein, when executed by the processor, the instructions further cause the processor to: in conjunction with performing one or more of the first plurality of sub-operations in accordance with the first communication model, perform the first sub-operation; in response to performing the first sub-operation, update the first dynamic token to reference a first plurality of changes made to the starting data at the first sub-operation; extract the first plurality of changes made to the starting data at the first sub-operation; determine whether the first plurality of changes is within the first tolerance of the first plurality of tolerances; in response to determining that the first plurality of changes is not within the first tolerance of the first plurality of tolerances, determine a communication anomaly associated with the first plurality of changes; determine one or more knowledge base commands configured to fix the communication anomaly; update the first plurality of configuration parameters to account for the one or more knowledge base commands; determine a second operational path to perform the first data exchange operation over a second path duration based on an updated version of the first plurality of configuration parameters; assign a second static token to the first data exchange operation based on the first communication level, the second static token referencing the first plurality of sub-operations to be performed on the second operational path over the second path duration; assign a second dynamic token to the first data exchange operation based on the first evaluation level, the second dynamic token referencing the starting data to be modified by the first plurality of sub-operations on the second operational path over the second path duration; train the first communication model to perform the first plurality of sub-operations of the second operational path using the second static token, the second dynamic token, the first plurality of tolerances, and the updated version of the first plurality of configuration parameters; and perform one or more of the first plurality of sub-operations in accordance with an updated version of the first communication model.
19. 19 . The non-transitory computer-readable medium of claim 18 , wherein, when executed by the processor, the instructions further cause the processor to: use the updated version of the first communication model to train one or more machine learning models.
20. 20 . The non-transitory computer-readable medium of claim 16 , wherein: the first communication level is an enriched communication level that comprises tracking inputs and outputs associated with the first data exchange operation from an origin point to a completion point in the first operational path.

Description

TECHNICAL FIELD The present disclosure relates generally to evaluating data exchange operations, and more specifically to a system and method to dynamically detect anomalies in data exchange operations. BACKGROUND In communication systems, data exchanges between two devices may involve multiple services. For example, a request may be processed, routed, and/or forwarded by multiple services between a first device and a second device. As the data is exchanged, the services may be configured to modify one or more portions of initial data included in the request. For example, a request to access power resources in a power station may comprise encrypted credentials that a first service in a communication path decrypts prior to forwarding a decrypted version of the request to a second service. In this case, the initial data comprising encrypted portions is modified to comprise decrypted portions. As the services modify the initial data, the services may introduce errors into the data exchange. Following the previous example, the decrypted data may be decrypted using an incorrect decryption process. Herein, the entirety of the data exchange operation may be compromised if errors are introduced by one or more services. SUMMARY OF THE DISCLOSURE In one or more embodiments, a system and method are configured to evaluate data exchange operations, and more specifically to dynamically detect anomalies in data exchange operations. In particular, the system may be configured to train a machine learning (ML) model to determine one or more anomalies and/or errors in processing performed by one or more services in one or more operational paths. In some embodiments, data exchanges between two or more devices may involve multiple services. For example, a request may be processed, routed, and/or forwarded by multiple services between a first device and a second device. As the data is sent from the first device to the second device, the services may be configured to modify one or more portions of the starting data included in the request. For example, a request to access communication resources in a base station may comprise signaling comprising a first format that a first service in the operational path may transform to a second format prior to forwarding a transformed version of the request to a second service. In this case, the starting data comprising the request in the first format is modified to comprise the second format. As the services modify the starting data, the system is configured to execute an ML algorithm to determine and address one or more anomalies introduced in the starting data. The services may be configured to inhibit, reduce, and/or eliminate anomalies and/or errors that may be introduced as part of modifications to the starting data. The system may be configured to train the ML model to determine preferred operational paths to complete one or more specific data exchange operations and evaluate inputs and/or outputs for every service along a specific operational path. The system may be configured to use one or more tokens referencing one or more aspects of the specific operational path, track possible modifications to the starting data as one or more services in the operational path modify the starting data, and correct any determined anomalies. In some embodiments, the actions and/or operations may be evaluated by one or more ML algorithms in accordance with the ML models. The ML models may be trained to understand and/or predict operations associated with specific anomalies in a specific operational path. The system may be configured to provide data exchange operation anomaly detection as a service using reservoir computing, multi-level static and dynamic tokens generated in association with one or more decentralized networks and one or more models, and generative artificial intelligence. In this regard, the system may be configured to find the anomalies (e.g., errors and/or issues) in an operational path. Herein, the system may be configured to use reservoir computing to detect one or more suitable operational paths for a specific data exchange operation. The system may be configured to dynamically determine specific services as part of one or more operational paths based on one or more configuration parameters and/or historical data. In some embodiments, static tokens may be generated to represent one or more expected modifications (e.g., changes) to starting data over an entirety of a data exchange operation while dynamic tokens may be generated to represent one or more expected modifications to the starting data as one or more sub-operations of the data exchange operations are performed by one or more services. For example, the static tokens may represent changes in the starting data at a communication level while the dynamic tokens may represent changes in the starting data at an evaluation level (e.g., at every hop in the operational path). The ML algorithm may be configured to create rules for ev