Masked 2D Vector Exchange for Private Data Aggregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In machine learning and other applications, aggregating information from multiple parties raises privacy concerns as central entities can identify which local party provides which information, compromising the privacy of local parties.
Innovation Solution
Implementing a privacy-preserving information exchange method using two-dimensional vectors with masking, where each local party stores values in a 2D vector with unique positions, and transmits masked vectors to an aggregator, allowing aggregation without revealing the originating party, and enabling decoding of the aggregated data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If local parties transmit their information to a central aggregator, then information aggregation efficiency is improved, but the aggregator can identify which local party provides which information, compromising privacy
Solution Approach 1:
The patent introduces masked vectors as an intermediary representation between local party data and the aggregator. Each local party's information is transformed into a masked vector that preserves the information content for aggregation purposes while removing identifying characteristics. The masking mechanism acts as a mediator that allows the aggregator to process and aggregate information efficiently without being able to trace it back to specific local parties, thus resolving the contradiction between aggregation efficiency and privacy preservation.
Solution Approach 2:
The patent transforms the original information from local parties by changing its parameters through masking operations. The information is converted from a state where it contains identifying features to a masked state where identifying features are removed or obfuscated. This parameter transformation allows the information to maintain its utility for aggregation while losing the characteristics that would compromise privacy, effectively resolving the technical contradiction.
2Object-affected harmful factors
If masking is applied to prevent the aggregator from decoding individual vectors, then privacy is preserved, but the ability to decode and utilize individual information is reduced
Solution Approach 1:
The masking mechanism applies partial obfuscation rather than complete encryption. The masking is designed to remove only the identifying portions of the information while preserving the substantive content needed for aggregation. This partial action approach ensures that privacy is protected to the necessary extent without unnecessarily compromising the utility and accessibility of the information for its intended purpose.
Solution Approach 2:
The masking process selectively changes parameters of the information representation. It transforms the data in a way that modifies identifying parameters while preserving parameters essential for aggregation and analysis. This selective parameter transformation maintains the balance between privacy protection and information accessibility, allowing the aggregator to utilize the information effectively without being able to identify individual sources.
Data Source
AI summary
Methods and system for privacy preserving information exchange in a network of electronic devices are disclosed. In one embodiment, a method is implemented in an electronic device to serve as a local party for information exchange between the local party and another electronic device to serve as an aggregator. The method includes storing a plurality of values in a 2D vector, where a first dimension of the 2D vector is based on the number of values, and where each position in the first dimension has one unique value. The method further includes transmitting the 2D vector to the aggregator with masking for the aggregator to prevent the aggregator from decoding the 2D vector, where aggregating the masked 2D vector with masked 2D vectors from other local parties allows decoding of the aggregated 2D vector.