Secure Data Similarity Detection Using Homomorphic Summaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for comparing private data while maintaining privacy are inefficient and vulnerable to manipulation, as they either require significant computational resources or allow parties to evade detection by manipulating data.
Innovation Solution
The use of partially homomorphic cryptographic schemes allows a system to securely compare private data by generating summaries without exposing plaintext, using a client-server setup where the client has access to the private key and the server performs homomorphic operations on ciphertext, ensuring that neither party can manipulate the results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exact similarity detection is used, then measurement precision is improved, but computing resources scale linearly with data object size
Solution Approach 1:
The data object is divided into multiple segments or blocks, and similarity detection is performed on each segment independently using parallel processing. This reduces the linear scaling burden by distributing the computational load across multiple processing units, maintaining exact similarity detection accuracy while improving computing resource efficiency through parallelization.
Solution Approach 2:
The system performs exact similarity detection on a selected subset of critical data segments rather than processing every single data point. This partial action approach maintains measurement precision for the most important portions of the data while reducing overall computing resource consumption by excluding less critical segments from exhaustive analysis.
2Productivity
If approximate similarity detection is used, then computing resources are conserved, but data privacy can be compromised through manipulation
Solution Approach 1:
Cryptographic hash values are pre-computed and embedded into the data structure before similarity detection occurs. This preliminary cryptographic preparation ensures that even when approximate methods are used, the embedded hash values provide a secure reference point that prevents manipulation, thereby maintaining detection security while allowing computing resource efficient approximate comparison methods to be used.
Solution Approach 2:
Cryptographic hash values serve as an intermediary between the original data and the similarity detection process. The hash values act as a trusted mediator that preserves data integrity - the approximate similarity detection operates on the hash values rather than raw data, conserving computing resources while the cryptographic nature of the hashes prevents manipulation and maintains detection security.
3Loss of information
If data is encrypted for privacy, then data privacy is improved, but similarity detection becomes more difficult
Solution Approach 1:
The system transforms the encryption approach by using deterministic cryptographic hash functions that produce consistent outputs for identical inputs. This parameter change in the cryptographic method allows similarity detection to work on the hash values - the encryption/ hashing provides privacy protection while the deterministic nature of the hash function enables similarity comparison without requiring decryption, thus maintaining data privacy while reducing detection complexity.
Data Source
AI summary
A computer-implemented method for securely detecting data similarities may include (1) identifying a private data object subject to comparison to determine a similarity with at least one potentially similar data object, (2) establishing a connection between a client system and a server that is not authorized to access the private data object, (3) generating a summary of the private data object by applying a summary generation protocol between the client system and the server to the private data object without exposing plaintext from the private data object to the server and without exposing a parameter of the summary generation protocol to the client system, and (4) computing the similarity between the private data object and the potentially similar data object by using the summary of the private data object and a summary of the potentially similar data object. Various other methods, systems, and computer-readable media are also disclosed.