Textual Data Masking for Secure Multiparty Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secure multiparty computation (sMPC) applications face challenges in protecting personal identifiable information (PII) during AI and ML data processing, as they often fail to effectively mask and preserve sensitive textual data, leading to potential privacy breaches.
Innovation Solution
The method involves data contributors masking their textual data with unique contributor mask values, which are then aggregated and shuffled by a broker, using both analyst and broker mask factors to ensure privacy, preventing third parties from accessing original data. This process ensures that only the contributor can decode their own masked results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is shared in encrypted form for secure multiparty computation, then data privacy is protected, but data utility for AI/ML analysis is reduced
Solution Approach 1:
The data is segmented into masked form where each contributor's data is transformed using their unique mask value. The segmentation separates the privacy protection function (masking) from the analysis function (aggregation on masked data), allowing both goals to be achieved simultaneously.
Solution Approach 2:
Masked data acts as an intermediary representation that preserves privacy while enabling analysis. The masked form serves as a mediator between the need for privacy protection and the need for data utility, allowing AI/ML algorithms to operate on data that cannot be directly traced to individual contributors.
2Reliability
If unique contributor mask values are applied to protect individual privacy, then data security is improved, but data aggregation for collective analysis becomes complex
Solution Approach 1:
The system changes the parameter representation by applying unique mask values to transform the data into a different parameter space. This parameter transformation allows individual privacy protection while enabling aggregation through mathematical operations on the masked values, resolving the complexity issue.
3Reliability
If masked data is aggregated and shuffled by a broker, then third-party access to original data is prevented, but the system requires trusted intermediaries increasing operational complexity
Solution Approach 1:
The broker serves as a trusted intermediary that facilitates secure data aggregation and shuffling. This intermediary enables third-party protection by ensuring that original data never leaves the contributors' control, while the broker's operational procedures maintain system simplicity through standardized masking and aggregation protocols.
Data Source
AI summary
Textual masking for multiparty computation is provided. The method comprises receiving masked input data from a number of contributors, wherein the input data from each contributor has a unique contributor mask value. A unique analyst mask factor is received for each contributor, computed by an analyst as a difference between a uniform analyst mask value and the contributor mask value. An API call is received from the analyst to aggregate the input data from the contributors. The respective analyst mask factors are added to the input data from the contributors, and the data is aggregated and shuffled. Computational results received from the analyst based on the aggregated input data are published. In response to API calls from the contributors, the analyst mask factors are removed from the computational results, wherein computational results received by each contributor are masked only by the respective contributor mask value.