Privacy-Preserving Multi-Party Analytics on Vertically Partitioned Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional solutions for privacy-preserving multi-party analytics on vertically partitioned data struggle to scale for generic analytical tasks, especially when more than two parties are involved, due to increased computational complexity and inability to handle big data effectively, while maintaining data privacy.
Innovation Solution
A secure protocol using a non-trusted mediator and public-key cryptography is employed to anonymize and compute a secure sum of data segments from multiple parties, allowing each party to maintain privacy by encrypting and shuffling data segments, ensuring collusion-resistant data anonymization and secure model parameter updates through gradient descent optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional privacy-preserving solutions are used for multi-party analytics, then data privacy is maintained, but the solutions cannot scale to generic analytical tasks with more than two parties due to increased computational complexity
Solution Approach 1:
The patent segments the data into vertically partitioned segments across multiple parties, where each party holds a portion of the features. This segmentation allows each party to perform local computations on their own data segment without needing to share the actual data, thereby maintaining privacy while enabling scalable multi-party analytics that can handle more than two parties without exponential complexity increases.
Solution Approach 2:
The patent introduces a mediator component that facilitates secure computation between parties. The mediator enables the coordination of encrypted computations and the aggregation of results without requiring parties to directly share their private data segments, thus reducing computational complexity while maintaining privacy guarantees in multi-party settings.
2Measurement precision
If data is shared across multiple parties to improve analysis quality, then data analysis quality improves, but privacy concerns prohibit the parties from sharing their data
Solution Approach 1:
The patent replaces the mechanical system of direct data sharing with a cryptographic system. Instead of physically or digitally sharing data segments between parties, the system uses encryption and secure computation protocols to allow analysis on the combined data while maintaining privacy. This substitution enables high-quality analytics without the harmful privacy exposure that would result from direct data sharing.
Solution Approach 2:
The patent changes the state of the data from plaintext to encrypted form, and changes the computation mode from direct processing to secure multiparty computation. By transforming data parameters and computation methods, the system achieves both high analysis quality (through access to multiple data segments) and privacy protection (through encryption and secure protocols).
Data Source
AI summary
Example computing devices described herein enable computation of a machine learning model on distributed multi-party data that is vertically partitioned, in a privacy preserving fashion. The computing device computes at a party a sum of local data owned by the party, wherein the local data is vertically partitioned into a plurality of data segments, each data segment representing a non-overlapping subset of data features; transforms a cost function of a data analytics task to a gradient descent function, wherein the cost function comprises a summation of a plurality of cost function values; anonymizes aggregated data shards received from a mediator; updating local model parameters based on the aggregated data shards; and performs privacy-preserving multi-party analytics on the vertically partitioned local data based on a learned global analytic model. It leverages a secure-sum protocol that provides strong security guarantees against collusion and prior-knowledge attacks.