A minimalist approach to Remote Attestation

Abstract

Embedded computing devices increasingly permeate many aspects of modern life: from medical to automotive, from building and factory automation to weapons, from critical infrastructures to home entertainment. Despite their specialized nature as well as limited resources and connectivity, these devices are now becoming an increasingly popular and attractive target for attacks, especially, malware infections. A number of approaches have been suggested to detect and/or mitigate such attacks. They vary greatly in terms of application generality and underlying assumptions. However, one common theme is the need for Remote Attestation, a distinct security service that allows a trusted party (verifier) to check the internal state of a remote untrusted embedded device (prover). Many prior methods assume some form of trusted hardware on the prover, which is not a good option for small and low-end embedded devices. To this end, we investigate the feasibility of Remote Attestation without trusted hardware. This paper provides a systematic treatment of Remote Attestation, starting with a precise definition of the desired service and proceeding to its systematic deconstruction into necessary and sufficient properties. Next, these are mapped into a minimal collection of hardware and software components that result in secure Remote Attestation. One distinguishing feature of this line of research is the need to prove (or, at least argue for) architectural minimality-an aspect rarely encountered in security research. This work also provides a promising platform for attaining more advanced security services and guarantees.

Attestation in Wireless Sensor Networks

“…Below we discuss each of them individually and highlight the design principles that must be followed to provide secure attestation. As stated before, we adopt much of previous analyses [Datta et al 2009;Coker et al 2011;Armknecht et al 2013;Li et al 2014;Francillon et al 2014] while centering the discussion around WSNs. CHALLENGE() This procedure is executed by the verifier.…”

“…The devices are commonly named after their roles: Verifier and Prover, respectively. The goal of the attestation procedure is to allow an honest, noncompromised, Prover to generate a response that assures the Verifier that the prover is in a legitimate state [Francillon et al 2014]. A compromised Prover will either generate invalid responses or will not be able to generate a valid response within an expected time limit.…”

“…Nonetheless, existing analyses are built under different assumptions. For example, [Datta et al 2009], [Coker et al 2011] and [Francillon et al 2014] assume that the verifier and the prover need to share some secret information, e.g., a cryptographic key. They presume that this secret is defined prior to network deployment and that there exists some hardware support preventing the adversary from accessing it.…”

“…They also claim that, contrary to some existing schemes, all memories of the prover must be attested and that doing so is quite challenging. [Francillon et al 2014] asserts that software-based techniques can be secure only if the prover and verifier communicate directly with no intermediate nodes, and thus cannot be used to perform attestation across a network. However, [Yang et al 2015] presents a delay-resilient software-based attestation mechanism capable of performing multi-hop attestation named Low-cost Remote Memory Attestation (LRMA).…”

Provably-Secure Remote Memory Attestation for Heap Overflow Protection

“…The works that is perhaps closest in spirit and application domain to ours is by Francillon et al [18] who address the problem of remote device attestation. Their approach is also based on provable security, but consider a significantly weaker model where the adversary not tamper or read parts of the internal memory of the device.…”