This paper presents a framework and implementation guidelines to set up nested compartmentalisation in constrained devices. All memory spaces are protected by the Memory Protection Unit (MPU). Current MPU-based systems offer efficient memory protection but are mostly tied to the fixed permission model provided by their operating system, kernel, hypervisor or by code instrumentation. New use cases evolve with the rise of the Internet of Things (IoT) ecosystems where software components could benefit from locally and dynamically established permissions. This includes a temporary nested subspace with restricted memory access rights. Our framework integrates subspace creation and management for runtime dynamic changes of the permission model for any level of abstraction. Global security policies of fixed permission models are reflected in the software architecture and the implementation of the framework. We also demonstrate the feasibility of providing nested compartmentalisation by showing how to leverage the MPU features.

IoT devices often operate unsupervised in ever-changing environments for several years. Therefore, they need to be updated on a regular basis. Current approaches for software updates on IoT, like the recent SUIT proposal, focus on granting integrity and confidentiality but do not analyze the content of the software update, especially the IoT application which is deployed to IoT devices. To this aim, in this paper, we present IoTAV, an automated software analysis framework for systematically verifying the security of the applications contained in software updates w.r.t. a given security policy. Our proposal can be adopted transparently by current IoT software updates workflows. We prove the viability of IoTAV by testing our methodology on a set of actual RIOT OS applications. Experimental results indicate that the approach is viable in terms of both reliability and performance, leading to the identification of 26 security policy violations in 31 real-world RIOT applications.