Random-Element Ciphertext Masking for Private Decryption Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in securely decrypting ciphertexts without allowing the server to learn or modify the plaintext, especially in scenarios involving sensitive data, and in efficiently utilizing shared encrypted models.
Innovation Solution
A processor masks ciphertexts using four random elements to generate masked ciphertexts, which are then decrypted by a server, and the masked plaintexts are unmasked using the same elements to ensure the server cannot learn or modify the data, while also enabling verification of unmodified plaintexts through multiplicative inverses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a server decrypts ciphertexts to provide decryption services, then decryption functionality is achieved, but the server may learn or modify the plaintext compromising privacy and security
Solution Approach 1:
The decryption process is segmented into two independent phases: (1) the server performs decryption on masked ciphertexts to obtain masked plaintexts without being able to learn or modify the original plaintext, and (2) the client unmaskes the masked plaintexts using random elements to recover the original plaintext. This segmentation ensures the server cannot access the actual plaintext while still providing decryption functionality.
Solution Approach 2:
Masked plaintexts serve as an intermediary between the server's decryption output and the client's final plaintext recovery. The masking mechanism using random elements acts as a mediator that prevents the server from learning or modifying the actual plaintext while allowing the decryption service to function. The client uses the random elements to unmask and recover the original plaintext.
2Reliability
If verification mechanisms are added to ensure plaintext integrity, then security is improved, but computational overhead and complexity increase
Solution Approach 1:
The masking of ciphertexts with random elements is performed in advance before decryption, which inherently creates a verification mechanism. The client can verify plaintext integrity by checking whether the unmasking process with the stored random elements successfully recovers the expected plaintext, without requiring separate complex verification protocols.
3Reliability
If encrypted models are used to maintain security, then privacy is preserved, but model evaluation and inference become more complex
Solution Approach 1:
The model evaluation process is segmented so that the server evaluates the encrypted model on masked ciphertexts and returns masked results, while the client performs the final unmasking and interpretation. This segmentation allows encrypted model usage without requiring the client to handle complex encrypted computations directly.
Data Source
AI summary
An example system includes a processor to mask a ciphertext using four random elements to generate masked ciphertexts. The processor can send the masked ciphertexts to a server device. The processor can receive masked plaintexts from the server device. The processor can unmask the masked plaintexts using the four random elements to generate unmasked plaintexts.