Genomic Data Privacy via Probabilistic Hashing and Deterministic Encryption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for de-identifying genomic data for epidemiological studies fail to balance privacy preservation with the need for efficient correlation analysis, as deterministic encryption methods render data uncorrelatable across patients, while probabilistic methods may not adequately protect patient anonymity.
Innovation Solution
A computer-implemented method that combines probabilistic irreversible hashing with deterministic reversible encryption to encode genomic data, creating a look-up table using a cryptogram as a key and locus information as a value, allowing for efficient correlation analysis while maintaining patient privacy by projecting data into reduced dimensions that are statistically improbable to recover outside the system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deterministic encryption methods are used to de-identify genomic data, then patient privacy is protected, but data becomes uncorrelatable across patients
Solution Approach 1:
The patent segments genomic data into two distinct portions: locus information (chromosome ID and position) and nucleotide sequence data. Each portion is processed differently - locus information undergoes probabilistic hashing while nucleotide data receives deterministic encryption. This segmentation allows the system to maintain privacy protection for identifiable locations while preserving correlatability through the hashed locus keys that can be matched across patients.
Solution Approach 2:
The patent transforms locus information into a different parameter space through probabilistic hashing, converting precise chromosome position data into reduced-dimensional hash values. This parameter transformation maintains the ability to correlate identical loci across patients while making it computationally infeasible to reverse-engineer the original positions, thus balancing privacy with correlatability.
2Loss of information
If probabilistic hashing is used to de-identify genomic data, then data correlatability is maintained, but patient anonymity may not be adequately protected
Solution Approach 1:
By separating locus information from nucleotide sequence data and applying different encryption strategies to each, the patent ensures that even if probabilistic hashing of locus information is broken, the deterministic encryption of nucleotide data provides an additional layer of anonymity protection. Conversely, if deterministic encryption is broken, the probabilistic nature of locus hashing still protects patient identity.
Solution Approach 2:
The patent creates a composite encryption scheme combining probabilistic hashing and deterministic encryption methods. This composite approach leverages the strengths of both methods - the correlatability of probabilistic hashing and the strong anonymity protection of deterministic encryption - while mitigating their individual weaknesses.
3Reliability
If genomic data is encrypted for privacy preservation, then patient anonymity is protected, but efficient correlation analysis becomes difficult
Solution Approach 1:
The patent performs preliminary probabilistic hashing on locus information before the correlation analysis begins. This pre-processing step creates compact hash keys that enable efficient matching and correlation across patients without requiring computation on the full encrypted genomic data during the analysis phase, thus maintaining both anonymity and analytical efficiency.
Solution Approach 2:
The patent creates hashed copies of locus information that serve as surrogate keys for correlation analysis. These hash copies retain the essential matching properties needed for correlation while containing no directly identifiable information, allowing efficient analysis on the copied data without compromising patient anonymity in the original data.
Data Source
AI summary
A method includes accessing genomic data of from a genomic database; generating, by a processor, a first hash by probabilistically and irreversibly encrypting a first portion of the genomic data encoding the first genomic sequence, the first hash projecting the first portion into reduced dimensions such that the first portion of the genomic data encoding the first genomic sequence becomes statistically improbable to recover outside the first processor; generating, by the processor, a first cryptogram by deterministically and reversibly encrypting a second portion of the genomic data encoding the first genomic sequence; generating, by the processor, a look-up table by using at least the first cryptogram as a key and the first portion of the genomic data encoding the first genomic sequence as the value, and transmitting data encoding the first hash and the first cryptogram to one or more processors that are different from the first processor.