Practical Privacy-Preserving Benchmarking

Abstract

Benchmarking is an important process for companies to stay competitive in today's markets. The basis for benchmarking are statistics of performance measures of a group of companies. The companies need to collaborate in order to compute these statistics.Protocols for privately computing statistics have been proposed in the literature. This paper designs, implements and evaluates a privacy-preserving benchmarking platform which is a central entity that offers a database of benchmark statistics to its customers. This is the first attempt at building a practical privacy-preserving benchmarking system and the first attempt at addressing all necessary trade-offs.The paper starts by designing a protocol that efficiently computes the statistics with constant cost per participant. The protocol uses central communication where customers only communicate with the central platform which facilitates a simple practical orchestration of the protocol. The protocols scale to realistic problem sizes due to the constant communication (and computation) cost per participant of the protocol.

Automatic Protocol Selection in Secure Two-Party Computations

“…Proofs for the protocols using homomorphic encryption can be found in [1,17,27]. For security of the mixed protocol we refer to Goldreich's composition theorem [18].…”

“…The intermediate language currently supports the following operations for which secure protocols are given in [1,17,27,33]. Some of these operations leverage the specific advantages of the respective protocol type, i.e., direct access to single bits and shift operations for garbled circuits or arithmetic operations for homomorphic encryption:…”

Improved Garbled Circuit Building Blocks and Applications to Auctions and Computing Minima

“…Our minimum selection protocol can also be used as a provably secure 5 replacement for the minimum selection protocol of [Ker08], which was used in the context of privacy-preserving benchmarking. In this scenario, mutually distrusting companies want to compare their key performance indicators (KPI) with the statistics of their peer group using an untrusted central server.…”

“…Our protocol for finding the nearest neighbor is a more efficient protocol for the special case k = 1. A simple protocol to select the minimum of homomorphically encrypted values based on the multiplicative hiding assumption was given in [Ker08] in the context of privacy-preserving benchmarking. However, multiplicative blinding reveals some information about the magnitude of the blinded value.…”

Selective access for supply chain management in the cloud

“…Data commonly collected in supply chains include time, location, and type of handling (e.g., packing, unpacking, receiving, or shipping). On one hand, combining these data from all the companies along a supply chain (not just predecessor and successor of each phase) enables or improves many economically attractive collaborative applications, including batch recalls [29], counterfeit detection [28], benchmarking and analytics [18]- [20], or estimated arrival forecasts [6]. On the other hand, information collected along the supply chain may be considered sensitive as it allows espionage on the business operations of the involved companies [12], [23].…”