Jonathan Cefalu, CEO and founder of NeuroBinder, presents the “Secure Hardware Architecture for Embedded Vision” tutorial at the May 2021 Embedded Vision Summit.
Security is a problem for every IoT system, but due to privacy concerns, security is particularly worrisome for embedded vision systems. In this talk, Cefalu covers how to design your embedded device so that the hardware architecture itself enforces strict guarantees about where visual data is able to flow.
In particular, Cefalu explores the idea of a “data diode” that uses hardware to enforce what parts of the system have access to video or images. This provides the highest level of protection against hacking or malware, as even if the device is completely hacked at the root level, the intruder will still be unable to access the visual data.
See here for a PDF of the slides.