Over the last years, security-kernels have played a promising role in reshaping the landscape of platform security on embedded devices. Security-kernels, such as separation kernels, enable constructing high-assurance mixed-criticality execution platforms on a small TCB, which enforces isolation between components. The reduced TCB  minimizes the system attack surface and facilitates the use of formal methods to ensure the kernel functional correctness and security.

In this thesis, we explore various aspects of building a provably secure separation kernel using virtualization technology. We show how the memory management subsystem can be virtualized to enforce isolation of system components. Virtualization is done using direct-paging that enables a guest software to manage its own memory configuration. We demonstrate the soundness of our approach by verifying that the high-level model of the system fulfills the desired security properties. Through refinement, we then propagate these properties (semi-)automatically to the machine-code of the virtualization mechanism.

Further, we show how a runtime monitor can be securely deployed alongside a Linux guest on a hypervisor to prevent code injection attacks targeting Linux. The monitor takes advantage of the provided separation to protect itself and to retain a complete view of the guest.

Separating components using a low-level software cannot by itself guarantee the system security. Indeed, current processors architecture involves features that can be utilized to violate the isolation of components. We present a new low-noise attack vector constructed by measuring caches effects which is capable of breaching isolation of components and invalidates the verification of a software that has been verified on a memory coherent model. To restore isolation, we provide several countermeasures and propose a methodology to repair the verification by including data-caches in the statement of the top-level security properties of the system.