[THESIS DEFENSE] Quentin Jayet - "Hardware-based elapsed-time attestation for embedded systems"

Abstract

Embedded systems, which are omnipresent in areas such as the Internet of Things, connected and autonomous vehicles, and industrial infrastructure, are subject to significant constraints in terms of resources, computing power, and energy consumption. Given their growing number and increasing interconnectivity, the issue of security and reliability of these networks is becoming essential. In this context, blockchain appears to be a promising solution. It allows the creation of a history of time-stamped events, shared and replicated in a distributed network where peers do not necessarily trust each other. This history is secured by a consensus protocol based on a proof mechanism that establishes trust. The proof of work mechanism, as used in Bitcoin, for example, relies on a very large number of calculations, which results in considerable energy consumption. This energy consumption is incompatible with the constraints of embedded systems. To address energy consumption issues, approaches based on measuring elapsed time have been proposed. These approaches involve replacing intensive calculations with random waiting, the validity of which is guaranteed by hardware security components, as is the case with proof of elapsed time. While this approach shows promise, it nevertheless has certain limitations: it was primarily designed for non-embedded environments dependent on a single manufacturer, and it suffers from vulnerabilities identified in the consensus protocols that use it. This thesis aims to study and design a proof mechanism based on elapsed time, specifically adapted to embedded systems and usable within a blockchain. The objective is to develop a low-power proof mechanism that can be used in the constrained devices, that is secure, independent of a manufacturer, and compatible with asynchronous operation. To this end, a new proof of hardware time mechanism is introduced. The security of this mechanism is studied, particularly under temperature attacks affecting the measurement of elapsed time. Finally, a theoretical model and an analysis of the probability of forks occurring in blockchains complete this study.

Jury members

Guests

 This thesis was completed at the laboratory CEA-LETI  [DRT/LETI/DSYS/SSSEC/LSES]