Retail item classification AI Engine in Trusty


We have an AI engine that can detect retail items on PoS terminal (via a camera).
We want to secure the AI model by putting it in Trusty, and calling it from the non-secure part of the OS.
Two questions:

  1. Is it a good idea to put the whole model (decision engine) in Trusty? Any link to design principles related to Trusty would be very helpful.
  2. Can someone please provide an example of same/similar work. If not possible/available, then an example of a complex application in Trusty would do. The sample application are fairly simple and rudimentary.

hello sine.language,

you may developing a Trusted Application (TA) that run with Trusty’s Trusted Execution Environment (TEE).
please also refer to the Architecture for the relationships among the Non-Secure Environment and Trusted Execution Environment.