Systems designed with measurement and attestation in mind are often layered, with the lower layers measuring the layers above them. Attestations of such systems, which we call layered attestations, must bundle together the results of a diverse set of application-specific measurements of various parts of the system. Some methods of layered attestation are more trustworthy than others especially in the presence of an adversary that can dynamically corrupt system components. It is therefore important for system designers to understand the trust consequences of different designs. This paper presents a formal framework for reasoning about layered attestations. We identify inference principles based on the causal effects of dynamic corruption, and we propose a method for bundling evidence that is robust to such corruptions.
展开▼