IEEE/MPAI 3301:2024

New IEEE Standard - Active - Draft.


The MPAI AI Framework (MPAI-AIF) Technical Specification (in the following also called MPAI-AIF V2) specifies the architecture, interfaces, protocols, and Application Programming Interfaces (API) of an AI Framework specially designed for execution of AI-based implementa-tions, but also suitable for mixed AI and traditional data processing workflows. Basic functionalities: 1. Independent of the Operating System. 2. Component-based modular architecture with specified interfaces. 3. Interfaces encapsulate Components to abstract them from the development environment. 4. Interface with the Store enabling access to validated Components. 5. Component can be Implemented as: 5.1. Software only, from MCUs to HPC. 5.2. Hardware only. 5.3. Hybrid hardware-software. 6. Component system features are: 6.1. Execution in local and distributed Zero-Trust architectures [28]. 6.2. Possibility to interact with other Implementations operating in proximity. 6.3. Direct support to Machine Learning functionalities. 7. The AIF can download an AIW whose identifier has been specified by the User Agent or by a configuration parameter. Secure functionalities: 1. The AIF provides access to the following Trusted Services: 1.1. A selected range of cyphering algorithms. 1.2. A basic attestation function. 1.3. Secure storage (RAM, internal/external flash, or internal/external/remote disk). 1.4. Certificate-based secure communication. 2. The AIF can execute only one AIW containing only one AIM. The AIM has the following features: 2.1. The AIM may be a Composite AIM. 2.2. The AIMs of the Composite AIM cannot access the Secure API. 3. The AIF Trusted Services may rely on hardware and OS security features already existing in the hardware and software of the environment in which the AIF is implemented.