# US 2026/0087167 A1 · Secure federated learning system for healthcare data management with privacy preservation and regulatory compliance

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0087167 A1
- Document ID: US20260087167A1
- Status: Published application
- Kind: A1
- Application number: 19/053,455
- Filed: 2025-02-14
- Published: 2026-03-26
- Earliest claimed: 2024-09-24
- Inventors: Sabira Arefin
- Web page: https://patentgrep.com/patent/US20260087167A1

## CPC

- G06F 21/6245: {Protecting personal data, e.g. for financial or medical purposes}
- G06F 21/554: {involving event detection and direct action}
- G06N 20/00: Machine learning
- G06F 2221/034: Test or assess a computer or a system
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The invention introduces a system and method for secure healthcare data management using federated learning, advanced encryption, and compliance monitoring. The system enables healthcare institutions to train machine learning models locally on sensitive data without transferring raw information, ensuring privacy and regulatory compliance. Model updates are encrypted using robust cryptographic techniques, such as AES and RSA, and transmitted securely to a central aggregator. Privacy-preserving protocols, including secure aggregation and differential privacy, ensure confidentiality during the creation of a global model that integrates insights from multiple institutions. The global model is validated locally, ensuring contextual relevance and continuous improvement. The system also incorporates real-time compliance monitoring to automate adherence to standards like HIPAA and GDPR, with detailed logging and corrective actions. Modular architecture supports seamless integration with existing infrastructures and flexible deployment options. This invention offers a scalable, secure, and privacy-preserving framework tailored to the complex demands of healthcare data security.

## Claims

### Claim 1

A method for federated learning in healthcare data security, comprising:
training, by at least one processor at a first healthcare institution, a local machine learning model on activity data stored within the institution, wherein raw data remains stored locally;
transmitting, by the at least one processor, model updates derived from the local machine learning model to a central aggregator;
aggregating, by at least one aggregator processor at the central aggregator, the model updates received from a plurality of healthcare institutions using a privacy-preserving protocol;
generating, by the aggregator processor, a global machine learning model based on the aggregated model updates; and
distributing, by the aggregator processor, the global machine learning model to the plurality of healthcare institutions.

### Claim 2 (depends on 1)

The method of claim 1, wherein the privacy-preserving protocol comprises secure aggregation to prevent reconstruction of raw data from the transmitted model updates.

### Claim 3 (depends on 1)

The method of claim 1, wherein differential privacy is applied to the model updates during the aggregating step.

### Claim 4 (depends on 1)

The method of claim 1, further comprising encrypting the model updates prior to transmission to the central aggregator using a symmetric or asymmetric encryption scheme.

### Claim 5 (depends on 4)

The method of claim 4, wherein the encryption scheme is selected from the group consisting of Advanced Encryption Standard, Rivest-Shamir-Adleman, elliptic curve cryptography, and combinations thereof.

### Claim 6 (depends on 1)

The method of claim 1, wherein the global machine learning model is configured to detect anomalies in healthcare data access patterns.

### Claim 7 (depends on 6)

The method of claim 6, wherein the anomalies detected by the global machine learning model are indicative of unauthorized access attempts or system misuse.

### Claim 8 (depends on 1)

The method of claim 1, wherein the local machine learning model is trained using activity data comprising user access logs, network traffic, and system interactions.

### Claim 9 (depends on 1)

The method of claim 1, further comprising evaluating the global machine learning model at each healthcare institution using locally stored validation data.

### Claim 10

A system for federated learning in healthcare data security, comprising:
a first processor at a first healthcare institution configured to:
train a local machine learning model on locally stored activity data while retaining raw data within the institution, and
transmit model updates derived from the local machine learning model to a central aggregator; and
a central aggregator comprising at least one aggregator processor configured to:
aggregate model updates received from a plurality of healthcare institutions using a privacy-preserving protocol,
generate a global machine learning model based on the aggregated model updates, and
distribute the global machine learning model to the plurality of healthcare institutions.

### Claim 11 (depends on 10)

The system of claim 10, wherein the privacy-preserving protocol comprises secure aggregation to prevent reconstruction of raw data from the transmitted model updates.

### Claim 12 (depends on 10)

The system of claim 10, wherein differential privacy is applied by the aggregator processor to the model updates during aggregation.

### Claim 13 (depends on 10)

The system of claim 10, wherein the first processor is configured to encrypt the model updates prior to transmission using a scheme selected from the group consisting of Advanced Encryption Standard, Rivest-Shamir-Adleman, elliptic curve cryptography, and combinations thereof.

### Claim 14 (depends on 10)

The system of claim 10, wherein the global machine learning model is configured to detect anomalies in healthcare data access patterns.

### Claim 15 (depends on 14)

The system of claim 14, wherein the anomalies detected by the global machine learning model are indicative of unauthorized access attempts or system misuse.

### Claim 16 (depends on 10)

The system of claim 10, further comprising a memory operatively coupled to the first processor, wherein the memory stores training data, model parameters, and encryption keys.

### Claim 17 (depends on 10)

The system of claim 10, wherein the local machine learning model is trained using user access logs, network traffic, and system interactions.

### Claim 18 (depends on 10)

The system of claim 10, wherein the aggregator processor redistributes the global machine learning model to the plurality of healthcare institutions via an encrypted communication channel.

### Claim 19

A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the processor to:
train a local machine learning model on activity data stored within a healthcare institution, wherein raw data remains stored locally;
transmit model updates derived from the local machine learning model to a central aggregator;
aggregate, at the central aggregator, model updates received from a plurality of healthcare institutions using a privacy-preserving protocol;
generate a global machine learning model based on the aggregated model updates; and
distribute the global machine learning model to the plurality of healthcare institutions.

### Claim 20 (depends on 19)

The non-transitory computer-readable medium of claim 19, wherein the instructions further cause the processor to encrypt the model updates using a scheme selected from the group consisting of Advanced Encryption Standard, Rivest-Shamir-Adleman, elliptic curve cryptography, and combinations thereof prior to transmission.
