Homomorphic Encryption Integrity Verification via Label Inversion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

There is a need to securely verify the integrity of application responses while protecting sensitive information from disclosure and ensuring that the applicant has not tampered with the results, particularly in environments of mutual distrust.

Innovation Solution

Implementing a method that uses fully homomorphic encryption (FHE) algorithms to encrypt applications, generating obfuscated application labels with trained machine learning models, and validating the integrity of decrypted responses by replicating and inverting labels, ensuring the client has not altered the results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If sensitive data is submitted to a data processor for processing, then the data processor can generate application responses, but the sensitive information may be disclosed to the data processor

Engineering Contradiction:
Improvedata privacy protectionVSAvoiddata processing capability
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

Homomorphic encryption serves as an intermediary mechanism that enables the data processor to work with encrypted data without decrypting it. The encryption scheme allows computational operations to be performed on ciphertexts, producing encrypted results that, when decrypted, match the results of operations on plaintexts. This intermediary layer protects sensitive information while maintaining processing capability.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If the data processor verifies the integrity of generated responses, then tampering can be detected, but the client's computational resources and time are increased

Engineering Contradiction:
Improveresponse integrity verificationVSAvoidvalidation processing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by generating multiple copies of the encrypted application labels and randomly inverting some of them before sending to the client. This preliminary preparation enables efficient integrity verification later, as the client can simply check whether the decrypted labels match the expected distribution pattern without performing complex computational validation.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The invention creates multiple copies of the encrypted application labels generated by the homomorphic computation. These copies are then randomly inverted and sorted to create a verification dataset. By working with copies rather than the original sensitive data, the system enables integrity verification while maintaining security and reducing the computational burden on the client.

Inventive Principle:
Principle #26Copying

3Reliability

If multiple copies of application labels are generated and inverted randomly, then integrity verification becomes possible, but the complexity of the processing algorithm increases

Engineering Contradiction:
Improveintegrity verification capabilityVSAvoidalgorithm processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The verification process is segmented into distinct, simple operations: generating multiple copies of labels, randomly inverting some copies, randomly sorting the inverted labels, and comparing the result against expected statistical properties. This segmentation breaks down the complex integrity verification task into manageable steps that are computationally efficient and easy to implement.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4270854B1Integrity protection for homomorphic computations
Publication Date: 2025.06.25 INTUIT INC
  • EP4270854B1 patent drawingFigure 1
  • EP4270854B1 patent drawingFigure 2
  • EP4270854B1 patent drawingFigure 3

AI summary

Systems and methods for securely verifying integrity of application responses are disclosed. One example method includes receiving, from a client, an application encrypted in accordance with a fully homomorphic encryption (FHE) algorithm, generating, with a trained machine learning model associated with the FHE algorithm, a plurality of first application labels, each first application label indicating a true or false response associated with the application, inverting a randomly selected portion of the plurality of first application labels, generating a first randomly sorted list including the plurality of first application labels, transmitting the first randomly sorted list to the client, receiving a first decrypted list from the client, performing a validation of at least the first decrypted list, the validation based at least in part on the plurality of first application labels, and in response to the validation being successful, providing the client with a response to the application. It is intended that, when published, the abstract be accompanied by Figure 2.