Property-based attestation in device swarms: a machine learning approach
Samuel Wedaj Kibret · 2022
Cyber-physical systems (CPS) and industrial Internet of things (IIoTs) are characterized by large numbers of tightly integrated heterogeneous components in a network. As such devices become imperative in mission-critical systems, their security is of immense concern. Toward this end, remote attestation (RA) schemes are widely used to verify the integrity of devices in IIoTs/CPS. However, such interactive mechanisms are vulnerable from verifier impersonation to replay and denial-of -service attacks. In addition to that, (binary) RA techniques are not resilient as any software upgrades would result in platform configuration and hash value changes. This necessitates repeated exchange of genuine hash values between collaborating devices. The problem is exacerbated because device swarms work in a group to attain a common goal, which is better identified by the system’s overall behavior. Training member devices using machine learning models and attesting them via property-based techniques help to increase the attained security and privacy performance of an Internet of things (IoT) swarm. In this chapter, to enhance the security guarantees of IoT networks and ensure swarm resiliency, we provide a comprehensive survey on the properties of IoT swarms. We also provide a concise overview of issues of IoT network attestation that arise from the fundamental nature of the underlying system, namely, the number of devices to be attested, the nature of member devices, and the property to be verified. We also describe a full prototype implementation and evaluate performance using OP-TEE, an Advanced RISC machine Trust- Zone-based open-source implementation of trusted execution environment. We also assess performance and analyze security for large swarms and show that the proposed approach is very effective and robust against various attacks.