الفهرس 
Chapter 1Only 14 pages are availabe for public view
 |  | from | 158 |  |  |
| | | from | 158 | | |
Abstract
The privacy aspect of state estimation algorithms has been drawing high research at- tention due to the necessity for a trustworthy private environment in cyber-physical systems. Such systems require estimation algorithms to estimate their unobservable states, which are either unmeasurable or subject to noise. These systems usually engage cloud computing services to aggregate the essential information from multiple spatially distributed nodes and then produce the desired estimates. The exchange of sensitive data among semi-honest parties raises privacy concerns, especially if there are coalitions between such parties that apply protocols correctly but keep records of their inputs, outputs, and intermediate computations to exploit in revealing private information of non-colluding parties.
This thesis presents two privacy-preserving cloud-based estimation protocols to estimate the hidden states of a system within two sensor setups: a typical one where sensors are spatially distributed and the other in which trustworthy sensors are grouped into sensor groups. The proposed protocols implement remote centralized and distributed estimations, where both employ different Kalman filter algorithms to produce optimal estimates using synchronously collected information from parties involved in the esti- mation process. To preserve privacy, the proposed protocols use partially homomorphic encryption to apply additional encryption only to sensitive information, i.e., measure- ments and estimates, without the covariances of random variables and other parameters in the system model.
The thesis also investigates privacy preservation in the proposed protocols. It proves that proposed protocols offer reasonable computational privacy guarantees against var- ious privacy-threatening coalitions based on the formal cryptographic definitions of the computational indistinguishability concept. Finally, we evaluate the proposed protocols to demonstrate their efficiency using actual data collected from a real test. We show that the proposed protocols provide satisfactory results with an acceptable level of esti- mation error, whereas the encryption process introduces neglectable encryption errors. This demonstrates that the proposed protocols preserves privacy while maintaining the estimation effectiveness.