Privacy-Preserving Distributed Optimising Control of Water Supply

Society relies on critical infrastructure, which is subject to an increasing threat of cyber attacks. This paper presents a method for designing distributed optimising controllers for critical infrastructure. The proposed method keeps local cost functions and constraints private in case of passive and eavesdropping attacks. The method is based on distributing a Model Predictive Control (MPC) problem using Alternating Direction Method of Multipliers (ADMM), which is made privacy-preserving through the use of Secure Multi-Party Computation (SMPC). The privacy-preserving controller is tested in a laboratory environment, emulating a Water Distribution Network (WDN), with real noise and disturbances, where it performs equally to a centralised model predictive controller. Thereby, the only cost of implementing the privacy-preserving controller is longer computation time and increased communication.