LegIoT: Ledgered Trust Management Platform for IoT

Abstract

We investigate and address the currently unsolved problem of trust establishment in large-scale Internet of Things (IoT) networks where heterogeneous devices and mutually mistrusting stakeholders are involved. We design, prototype and evaluate LegIoT, a novel, probabilistic trust management system that enables secure, dynamic and flexible (yet inexpensive) trust relationships in large IoT networks. The core component of LegIoT is a novel graph-based scheme that allows network devices (graph nodes) to re-use the already existing trust associations (graph edges) very efficiently; thus, significantly reducing the number of individually conducted trust assessments. Since no central trusted third party exists, LegIoT leverages Distributed Ledger Technology (DLT) to create and manage the trust relation graph in a decentralized manner. The trust assessment among devices can be instantiated by any appropriate assessment technique, for which we focus on remote attestation (integrity verification) in this paper. We prototyped LegIoT for Hyperledger Sawtooth and demonstrated through evaluation that the number of trust assessments in the network can be significantly reduced – e.g., by a factor of 20 for a network of 400 nodes and factor 5 for 1000 nodes.

A Attestation Schemes and Trust Scope

Remote attestation originally came to prominence as a feature of the TPM  [41], the standard defined by the Trusted Computing Group (TCG)  [48]. Many approaches to attestation have been developed, which differ in underlying requirements and security guarantees provided. Generally, they provide different levels of resilience, which refers to the general robustness of the underlying architecture against compromise. In the following we discuss attestation approaches of four categories, and suggest what resilience level and trust scope they can provide.

Hardware-Based Architectures. Include strong cryptographic co-processors like TPMs  [41]. A different approach are Trusted Execution Environments (TEEs) that use an isolated processing environment  [43]. Usually, they offer complex attestation mechanisms with arbitrary cryptographic functionality. Since cryptographic co-processors are well studied and strongly protected, hardware-based architectures generally have a high resilience. Thus, this architecture is able to attest other devices and create functional as well as referral trust.

Hybrid Architectures. Generally include minimal security features like Read Only Memory (ROM) and Memory Protection Unit (MPU) for secure storage [24]. Generally, hybrid schemes such as SMART  [20] and TrustLite  [32] attest a defined area of code only. Their limitations are less significant compared to software-based attestation schemes. Thus, their resilience is considered to be medium. If the attested code contains the segment that handles device functionality, functional trust is gained. In contrast, referral trust requires the attestation component of the prover to be attested.

Software-Based Attestation. Generally, secure co-processors are not available on low-end embedded devices due to minimal cost requirements. Thus, purely software-based approaches were developed  [45]. They do not assume any secrets on the prover’s device, since there is no secure storage available at the prover side. Instead, these schemes are based on using side-channel information to decide whether an attestation result is valid. However, this approach poses many assumptions on the network topology and adversarial capabilities. For instance, the verifier needs to have direct communication with the prover with no intermediate hops  [3]. We consider resilience of this attestation type as low because the potential attack surface is comparatively high. As attestation statements made by such attestations about other parties cannot be trusted, they can only provide functional trust.

Control-Flow Attestation is a relatively recent development in the attestation landscape  [2]. Static attestation, to which previously discussed attestation categories belong to, is not able to capture misbehavior of software during runtime. This is where runtime attestation comes into play by monitoring an application’s control flow and detecting all deviations from the expected flow (documented in the security policy). This approach enables the highest trust guarantees of all attestation schemes. Runtime attestation schemes like DIAT  [4] offer a very high resilience because they also protect against runtime adversaries, and thus can provide both referral and functional trust.