Open Se Cura is to develop a set of open source design tools and IP libraries
Recently Google unveiled through a blog post, the launch of a new project, which is named Open Se Cura, whose objective is simplify the creation of secure chips designed to solve problems related to el machine learning and artificial intelligence.
It is mentioned that among the application areas of Open Se Cura, are specialized chips that require a special level of protection and confirmation of the absence of failures. For example, the platform can be used in machine learning products related to the processing of sensitive information, such as systems for recognizing people and processing voice recordings.
Formerly known internally as Project Sparrow, Project Open Se Cura is a testament to our commitment to open source development. Our goal with Open Se Cura is to develop a set of open source design tools and IP libraries that will accelerate the development of complete systems with machine learning workloads through co-design and development. This will allow us to better focus system designs around security, efficiency and scalability, powering the next generation of AI experiences.
About Open Se Cura
Open Se Cura has under its hood, the CantripOS operating system, while on the hardware side, it is based on the OpenTitan platform and a processor core based on the architecture RISC-V.
The CantripOS operating system It is based on the seL4 microkernel, on which a system environment runs written in Rust language. In RISC-V systems, a mathematical proof of reliability is provided for the seL4 microkernel, indicating that the code fully meets the specifications specified in the formal language. The seL4 architecture stands out by moving parts for managing kernel resources into user space and applying the same access control tools for those resources as for user resources.
The microkernel does not provide abstractions high-level out-of-the-box to manage files, processes, network connections, and the like; instead, it provides only minimal mechanisms to control access to physical address space, interrupts, and processor resources.
High-level abstractions and controllers to interact with hardware are implemented separately on top of the microkernel in the form of user-level tasks. The access of these tasks to the resources available to the microkernel is organized by defining rules.
All components of the operating system, except for the microkernel, are originally written in Rust using safe programming techniques that minimize errors when working with memory. Among other things, an application loader in the seL4 environment, system services, an application development framework, an API for accessing system calls, a process manager, and a dynamic memory allocation mechanism are written in Rust.
The build is verified using CAmkES, which is developed by the seL4 project. To develop end applications that can be dynamically loaded by system services, it is proposed to use the AmbiML SDK and the IREE (Intermediate Representation Execution Environment) toolkit to run machine learning models. Rust system components and services are developed using Cantrip frameworks.
Finally, it is worth mentioning that in the development of the project Not only Google participated, as was also involved in the development of tools and infrastructure elements, the non-profit organization lowRISC, which supervises the development of a free microprocessor based on the RISC-V architecture, as well as the companies Antmicro and VeriSilicon.
The processor developed by lowRISC was used as the core to build trusted hardware components (Root of Trust). Antmicro provided the project with a Renode simulator that allows testing CantripOS and the seL4 microkernel without real hardware. VeriSilicon shared its experience in the field of chip creation and BSP (Board Support Package) development.
If you are interested in knowing more about it, you should know that the project developments, including the source code of system services and RTL schemes, They are distributed under the Apache 2.0 license.
Source: https://opensource.googleblog.com