Interruptible Remote Attestation of Low-end IoT Microcontrollers via Performance Counters

Abstract

Remote attestation is a method used in distributed systems to detect integrity violations on a target device (prover) through a challenge-response protocol initiated by a verifier device. The prover calculates a hash of its memory, which is compared to a known good state hash by the verifier. We propose a novel technique, called Counters Help Against Roving Malware (CHARM), which uses hardware performance counters on the prover's side and machine learning on the verifier's side to make interruptible remote attestation feasible, even for constrained microcontrollers. We will demonstrate the effectiveness of various machine learning tools and data manipulation techniques on prediction accuracy in a variety of scenarios.