Modern computing systems face security challenges that span multiple layers of abstraction, from hardware and firmware in embedded devices to the machine learning models that increasingly govern their behavior. As embedded systems become more intelligent and ML models are deployed in safety-critical environments such as autonomous vehicles, satellite systems, and industrial control, the attack surface grows in complexity. Adversarial manipulation of ML inputs, firmware-level vulnerabilities, and side-channel leakage each represent significant threats individually, yet their intersection remains underexplored.

This seminar brings together researchers from the Embedded Systems Security (EMSEC) and Adversarial Machine Learning groups at CISPA Helmholtz Center for Information Security to explore this intersection. Topics include the robustness of ML models deployed on embedded platforms, firmware security for systems that rely on learned components, adversarial attacks on sensor inputs and communication protocols, and the development of defensive mechanisms that account for both system-level and model-level threats. The seminar aims to identify new collaborative research directions and produce a shared roadmap for securing intelligent embedded systems.

Creative Commons BY 4.0

Ali Abbasi and Lea Schönherr