Privacy-preserving benchmarking with interval statistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing benchmarking systems leak confidential information about participating enterprises, making them reluctant to share sensitive production key performance indicators (KPIs), as rank-based statistical measures can reveal private data, and finding a trusted third party for privacy-preserving benchmarking is challenging.
Innovation Solution
A privacy-preserving benchmarking protocol using homomorphic encryption and semi-homomorphic encryption schemes that computes interval statistics instead of exact values, allowing participants to learn their performance relative to others without revealing individual KPIs, by employing techniques like oblivious transfer, summation, rank computation, and decryption, ensuring that only interval-based composite statistics are shared.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rank-based statistical measures (median, maximum, quartiles) are used for benchmarking, then performance evaluation accuracy is improved, but information leakage of confidential KPIs increases
Solution Approach 1:
The patent introduces an intermediary mechanism (secure multi-party computation protocol) that acts as a mediator between participating enterprises. This intermediary enables the computation of rank-based statistical measures without any single party having access to the raw confidential KPI data, thus preserving measurement precision while preventing information leakage.
Solution Approach 2:
The patent segments the computation process into multiple independent steps performed by different parties. Instead of one party having full access to all KPIs, the computation is divided into encrypted comparisons, rank determination, and statistical calculation phases, where each party only processes encrypted data or intermediate results, preventing any single entity from accessing confidential information.
2Reliability
If a trusted third party is used to conduct privacy-preserving benchmarking, then data privacy is improved, but system complexity and trust requirements increase
Solution Approach 1:
The patent implements a self-service mechanism where participating enterprises themselves perform the privacy-preserving computation through standardized protocols. Each enterprise runs local computation on encrypted data and contributes to the collective result without requiring an external trusted third party, thereby reducing system complexity and eliminating trust requirements while maintaining strong privacy protection.
3Measurement precision
If exact KPI values are shared for benchmarking, then statistical accuracy is improved, but enterprise privacy and security are compromised
Solution Approach 1:
The patent changes the parameter representation from exact KPI values to encrypted representations and rank-based statistics. By transforming the data into different parameter forms (encrypted values, ranks, intervals), the system maintains statistical accuracy for benchmarking purposes while the transformed parameters cannot be reverse-engineered to reveal the original confidential KPI values.
Data Source
AI summary
Disclosed herein are computer-implemented method, system, and computer-program product (computer-readable storage medium) embodiments for benchmarking with statistics in a way that reduces leakage, preserving privacy of participants and secrecy of participant data. An embodiment includes receiving a plurality of encrypted values and computing a composite statistic corresponding to at least a subset of the plurality of encrypted values. An embodiment may further include outputting the at least one composite statistic. The composite statistic may be calculated to be distinct from any encrypted value of the plurality of encrypted values, thereby preserving privacy. Further embodiments may also include generating a comparison between the composite statistic and a given encrypted value of the plurality of encrypted values, as well as outputting a result of the comparison. In some embodiments, encrypted values may be encrypted using at least one encryption key, for example, according to a homomorphic or semi-homomorphic encryption scheme.