Secure Control in the Cloud Using Multiparty Computation

Student thesis: Master thesis (including HD thesis)

  • Andrea Victoria Tram Løvemærke
4. term, Control and Automation, Master (Master Programme)
The society of today and even more so in the future relies on data to reduce resource consumption, wastage and overall costs. Technologies have been developed such that devices across different functionalities can communicate and make independent decisions to optimize their individual function. Online control algorithms are needed to achieve this kind of linked optimization and automation. However, online servers are not guaranteed to be trustworthy, leading to great risk when data is released to third parties in exchange for computed control actions. It is vital that privacy is preserved and data is not compromised through acts of terror, to ensure safety in today’s data-driven society.
This thesis is devoted to securing data within cloud computed control. Multiparty computation using secret sharing is investigated as a secure approach where private data is split into shares. Computing on shares can be performed on online servers with minimum risk as the shares individually do not disclose information about the private data. The data owner can reconstruct the private result of the secure multiparty computations with no other party gaining knowledge of the result.
It is investigated how unconstrained and equality constrained model predictive control problems can be securely solved in the cloud. The solution utilizes Gaussian elimination without pivoting and has proved to be useful. The thesis concludes that further research must be done within this field prior to become suitable for real life control problems. It does, however, show tremendous potential to becoming epochal for future online control applications.
LanguageEnglish
Publication date14 Jun 2019
Number of pages61
ID: 305803017