Software containerization technologies for resource-limited devices enable multi-tenant microcontrollers, which allow running multiple applications with different permission levels. However, most solutions lack runtime fine-grained container permissions and container scheduling.
In this presentation, we first introduce CS4WAMR and TinyContainer, two software container solutions running on RIOT OS and using WebAssembly, which allow configurable permissions and scheduling for the container.
Then, we focus of how we used software container solutions to enable the dynamic deployment and secure inference of Tiny Machine Learning (TinyML) models on microcontrollers. We notably compare two implementations: one delegating TinyML inference to the host and one running the inference entirely in the WebAssembly container.