Neural Network Intrusion Detection System Patent Grant

ChangeBridge: Patent Grants - Networking (H04L)

Published March 31st, 2026

Detected March 31st, 2026

Summary

The USPTO granted Bank of America Corporation Patent US12592955B2 for a neural network-based network intrusion detection system. The system intercepts and analyzes network requests, grouping suspicious requests by geolocation, evaluates request rates against thresholds, and uses neural networks to classify malicious activity. The patent, with 20 claims, was filed September 22, 2023.

View original document View source feed page

What changed

The USPTO granted Bank of America Corporation Patent US12592955B2 for a system and method of network intrusion detection using neural networks. The invention covers a multi-stage filtering process: intercepting requests, identifying authenticated versus suspicious requests, grouping suspicious requests by geolocation, analyzing request rates against thresholds, evaluating parameter values, and using a neural network to classify requests as legitimate or malicious. Three inventors (Ngoc Anh Tran, Manimaran Sundaravel, and Maneesh Kumar Sethia) are credited with the 20-claim patent.

This is a patent grant notice, not a regulatory requirement. Financial institutions, technology companies, and cybersecurity teams developing network security solutions should note this patent for freedom-to-operate considerations. Organizations implementing similar intrusion detection methods may want to review licensing terms with Bank of America. There is no compliance deadline or regulatory action required.

Source document (simplified)

← USPTO Patent Grants

System and method for network intrusion detection using a neural network implemented by a local computing system

Grant US12592955B2 Kind: B2 Mar 31, 2026

Assignee

Bank of America Corporation

Inventors

Ngoc Anh Tran, Manimaran Sundaravel, Maneesh Kumar Sethia

Abstract

A method includes intercepting requests. The requests are analyzed to identify authenticated requests. Remaining requests are identified as suspicious requests. The suspicious requests are grouped into request groups based on respective geolocation information. A first rate of requests is determined for a first request group. In response to determining that the first rate of requests is less than or equal to a request rate threshold, parameters of a first suspicious request of the first request group are analyzed to determining values of the parameters. In response to determining that the values of the parameters are not within respective acceptable parameter value ranges, the first suspicious request is analyzed using a neural network to identify if the first suspicious request is legitimate or malicious. In response to identifying that the first suspicious request is malicious, a first notification indicating that the first suspicious request is identified as malicious is send.