Standardized AI Architectures for Secure TinyML

Learn about the challenges and solutions for implementing machine learning on highly constrained edge devices in this 56-minute technical talk by Andrea Basso from EPFL and Stanford University. Explore how TinyML enables complex ML tasks to run on microcontrollers with limited resources, offering advantages like reduced bandwidth, lower energy consumption, enhanced privacy, and improved scalability. Dive into the critical security considerations unique to small devices deployed in the field, including hardware limitations and physical vulnerabilities. Discover how the MPAI-AIF framework and IEEE P3301 standard provide practical solutions for secure TinyML implementation on microcontrollers, addressing key challenges in memory constraints, processor performance, and power management.