Google presented Project Open SeCura, an open-source initiative that focuses on machine learning and artificial intelligence. The project includes the Cantripos operating system and hardware based on the Opentitan platform and RISC-V architecture. Originally developed as Sparrow and KataOS, the project’s components, including system services and rtl schemes, are now open under the Apache 2.0 license (source).
Cantripos is built upon the microkernel sel4 and uses a system environment written in Rust. Mathematical evidence of reliability demonstrates that the code complies with formal language specifications on RISC-V Systems for Micro-Sel4. SEL4 architecture facilitates resource management in both the user space and application space, ensuring equal access to resources (source).
The microkernel sel4 provides minimal mechanisms for managing access to physical target space, interruptions, and processor resources. High-level abstractions and drivers for managing files, processes, network connections, etc., are implemented separately as tasks in the user space. Rules define the access of these tasks to microkernel resources (source).
All OS components, except for the microkernel sel4, are originally written in Rust using safe programming techniques. Rust enables the development of applications loader, system services, application development framework, API for system calls, process manager, and a dynamic memory distribution mechanism (source).
The project leverages the instrumentation of camkes from the Sel4 project. To develop final applications that can dynamically load system services, the sdk ambiml and tools IREE (Intermediate Representation EXECUTION ENVIRONMENT) are proposed. These tools facilitate the deployment of machine learning models (source).