Watch a 49-minute conference talk from tinyML Asia 2022 exploring security vulnerabilities in TinyML systems. Delve into the unique challenges of protecting AI models deployed on tiny embedded devices, which differ significantly from traditional AI systems. Learn about various attack vectors including model theft, evasion, data poisoning, and membership inference, particularly critical as these devices often operate in unmonitored environments. Understand the kill-chain methodology for potential attacks through practical use-cases, while discovering defense mechanisms to protect TinyML applications in critical sectors like healthcare, smart factories, and surveillance. Gain insights from Bosch AI Shield's Senior Research Engineer on why traditional embedded security approaches may not suffice for securing AI on tiny devices, and explore the intersection of neuromorphic computing, on-device learning, and security considerations in the rapidly evolving field of TinyML.