Towards encrypted cloud-based control-as-a-service
| dc.contributor.advisor | Schulze Darup, Moritz | |
| dc.contributor.author | Schlüter, Nils | |
| dc.contributor.referee | Ferrari, Riccardo | |
| dc.date.accepted | 2024-11-07 | |
| dc.date.accessioned | 2025-05-30T04:58:28Z | |
| dc.date.available | 2025-05-30T04:58:28Z | |
| dc.date.issued | 2024 | |
| dc.description.abstract | The increasing interconnectivity of control systems paves the way for exciting new opportunities, such as cloud-based control services. At the same time, connected control systems are susceptible to cyberattacks and data leakage. Therefore, this dissertation aims to enhance the security of cyberphysical systems by ensuring the confidentiality of control data during its evaluation. % on external platforms. To achieve this, advanced cryptographic techniques such as homomorphic encryption and secure multi-party computation are utilized, which enable computations on encrypted data. However, the development of novel encrypted controllers results in unique interdisciplinary challenges that necessitate a nuanced approach, blending elements of cryptography and control theory. The first section of this thesis is devoted to iterative control algorithms. First, cloud-based dynamic controllers are considered. These are based on a linear iteration, which can lead to a numerical overflow in an encrypted evaluation. Our system-theoretic analysis reveals that a specific subset of dynamic controllers, which can be efficiently encrypted, is not affected by this problem. Second, privacy in consensus problems of multi-agent systems is explored. In the related algorithms, we address iterations and a mechanism to control the data access. The subsequent section of this thesis focuses on controllers that use non-po\-ly\-no\-mi\-al functions, where model predictive control is a prominent example. Encrypted implementations of these functions are often inefficient. Our novel approach uses max-out neural networks, which form a versatile basis to approximate any continuous function. Moreover, they are well-aligned with primitives from secure multi-party computation, resulting in an efficient implementation. The final section conducts a security analysis of random affine transformations. Although these transformations have gained significant popularity, they currently do not offer a technically sound security guarantee. The aforementioned results are numerically validated and illustrated with relevant examples. Although significant progress has been made, further efficiency improvements are paramount for the practical application of encrypted control. | en |
| dc.identifier.uri | http://hdl.handle.net/2003/43716 | |
| dc.identifier.uri | http://dx.doi.org/10.17877/DE290R-25490 | |
| dc.language.iso | en | |
| dc.subject | Encrypted control | en |
| dc.subject | Control-as-a-service | en |
| dc.subject | Homomorphic encryption | en |
| dc.subject | Secure multi-party computation | en |
| dc.subject.ddc | 620 | |
| dc.subject.rswk | Cyber-physisches System | de |
| dc.subject.rswk | Kryptologie | de |
| dc.subject.rswk | Computersicherheit | de |
| dc.subject.rswk | Mehrparteienprotokoll | de |
| dc.subject.rswk | Regelungstechnik | de |
| dc.subject.rswk | Numerisches Verfahren | de |
| dc.title | Towards encrypted cloud-based control-as-a-service | en |
| dc.type | Text | |
| dc.type.publicationtype | PhDThesis | |
| dcterms.accessRights | open access | |
| eldorado.secondarypublication | false |