Masterarbeit
Generic Secure Multi-Party Computation in Centralised Cloud-based Environments
Completion
2018/09
Research Area
Intelligent Information Management
Students
Advisers
Description
Comparing business KPIs with other market participants through
benchmarking as well as joint calculations of generic arithmetic functions is a means for
companies to optimize costs. Those collaborative optimizations require data of all
participating actors that might include business secrets, and therefore must be kept
private in many cases. This demonstrates the demand for privacy-preserving collaborative
optimisation techniques. Over the last decades, a variety of mechanisms and protocols that
enable privacy-preserving collaborative computations have been presented such as trusted
third party (TTP) approaches or secure multi-party computation (MPC). The idea of secure
MPC is to emulate a TTP by jointly computing a public function. Such a protocol is secure
in the sense that each participant only knows its own input, the computation’s
output, and what can be inferred from that.
Existing solutions for
privacy-preserving benchmarking compute only a fixed set of arithmetic functions. In
contrast, generic secure multi-party computation systems like FairplayMP enable the secure
computation of arbitrary functions. However, these usually follow a decentralised
communication scheme. The main objective of the master’s thesis therefore is (1)
to design, implement, and evaluate a generic secure computation system that can compute
arbitrary functions in a centralised communication scheme and (2) to determine the class
of arithmetic functions that the selected secure computation mechanism can compute
feasibly in practice.
To achieve this, the thesis will encompass
mechanisms for compiling a given arbitrary function into a secure MPC protocol as well as
for compiling this protocol into a runnable cloud-based implementation. This includes
literature research of the current research state with a focus on suitable protocols for
privacy-preserving addition and multiplication. The chosen mechanisms will be implemented
in a proof of concept prototype, which will be evaluated regarding security, performance,
and other criteria.
