Two-Server Sparse Histogram Protocol Using Homomorphic Encryption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for computing sparse, differentially private histograms in the two-server model of secure multi-party computation face challenges in achieving computational efficiency and scalability, particularly due to high communication and computation overheads, and the need for trusted curators or central aggregators.
Innovation Solution
The proposed solution involves a protocol where client devices homomorphically encrypt data entries and transmit them to two non-colluding servers for collaborative aggregation, using public and private homomorphic encryption keys to ensure privacy and security, with techniques like oblivious pseudorandom functions and dummy contributions to reduce communication and computation costs, independent of the domain size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If distributed MPC protocols are used to compute aggregate statistics, then privacy guarantees are improved, but computation and communication overhead increase significantly
Solution Approach 1:
The patent segments the computation task by introducing a trusted shuffler that separates the aggregation process into distinct phases: clients send encrypted data to the shuffler, the shuffler permutes and forwards to servers, and servers compute histograms. This segmentation allows distributed privacy-preserving computation while reducing the computational burden on individual participants compared to fully distributed MPC.
Solution Approach 2:
The patent introduces a trusted shuffler as an intermediary component that receives encrypted data from clients, performs random permutation, and forwards to servers. This intermediary enables the system to achieve privacy guarantees without requiring all clients to participate in complex multi-round MPC protocols, thereby improving computation efficiency while maintaining security.
2Reliability
If fully distributed MPC is implemented across all user devices, then privacy is enhanced, but scalability deteriorates for large user populations
Solution Approach 1:
The system segments the user population into clients who send encrypted data and servers who perform computation, with a shuffler managing the intermediate step. This segmentation allows the system to scale to large user populations by avoiding the need for every user device to participate in complex multi-party computation, while still providing strong privacy guarantees through encryption and permutation.
Solution Approach 2:
The trusted shuffler acts as an intermediary that enables scalability by centralizing the permutation operation. Instead of requiring all clients to engage in distributed MPC, the shuffler collects encrypted data, permutes it centrally, and distributes to servers, making the system scalable to billions of users while maintaining privacy through the cryptographic guarantees of the underlying protocols.
3Measurement precision
If central DP mechanisms are used with a trusted curator, then accuracy is improved, but trust requirements increase
Solution Approach 1:
The patent segments the trusted curator role into two separate components: a trusted shuffler that handles permutation and servers that handle histogram computation. This segmentation allows the system to maintain accuracy through proper implementation of central DP mechanisms while distributing trust across multiple components rather than requiring a single all-powerful curator, thereby reducing the complexity of the trust model.
Solution Approach 2:
The trusted shuffler serves as an intermediary between clients and servers, performing the permutation operation that enables central DP without requiring a single trusted curator to have access to all raw data. The shuffler's role is limited to cryptographic operations and permutation, reducing the trust requirements compared to a traditional central curator while maintaining accuracy through proper noise addition and histogram computation.
4Adaptability or versatility
If domain size increases, then histogram coverage is improved, but communication and computation costs increase
Solution Approach 1:
The patent applies local quality by having servers compute histograms locally on encrypted data without needing to understand or process the full domain space. Each server receives encrypted data, performs local histogram computation based on the encrypted values, and sends results to the trusted curator. This approach allows the system to handle large domain sizes efficiently because computation is performed locally on the actual data distribution rather than over the entire domain space.
Data Source
AI summary
Provided are systems and methods for the computation of sparse, (ε, δ)-differentially private (DP) histograms in the two-server model of secure multi-party computation (MPC). Example protocols enable two semi-honest non-colluding servers to compute histograms over the data held by multiple users, while only learning a private view of the data.