CPS-IoT Week 2026 Keynotes

---

Keynote 1: Sensing What Matters: From Human Physiology to Trustworthy Media

Date: May 12, 2026, 9 am - 10 am

Speaker: Xia Zhou, Columbia University
Chair: TBD

Abstract: We live in a world where both our bodies and environments are sensed at unprecedented scale and granularity. However, the trustworthiness of sensed data is increasingly in question. Physiological sensing suffers from motion noise, user variability, and placement uncertainty, while audiovisual media is challenged by the rise of deepfakes and AI-generated content that blur the boundary between real and fake. The fundamental tension is that sensing is no longer just about acquiring signals — it is about establishing trust.
In this talk I present our efforts to make sensing systems trustworthy by design. For human sensing, we explore everyday fabrics as a ubiquitous, continuous sensing medium. Through hardware-software co-design, we address real-world challenges such as motion noise and user diversity, enabling applications ranging from physical motion sensing to physiological monitoring for kangaroo mother care and sleep.
On sensing the physical world, we investigate embedding verifiable physical signatures directly into the environment. I will present our design of imperceptible spatiotemporal light signatures that can be projected into a scene and embedded into any video recordings, enabling verification of live speech videos without requiring compliance from recording parties. I will also discuss our ongoing effort of embedding real-time, unforgeable, and robust audio watermarks into live speech audios, enabling verification of audio integrity.
Together, these efforts point toward a sensing paradigm where trust is not inferred after the fact, but physically embedded into signals and environments from which data is captured.

Keynote 2: Performance Isolation in Embedded Systems: A HW/SW Co-design Perspective

Date: May 13, 2026, 9 am - 10 am

Speaker: Rodolfo Pellizzoni, University of Waterloo
Chair: TBD

Abstract: Modern embedded systems are increasingly deployed on heterogeneous Multiprocessor Systems-on-Chip (MPSoCs), where general purpose instruction processors interact with accelerators through complex memory hierarchies. Applications with different safety integrity levels are integrated in virtualization environments. In addition to providing logical isolation, the environment must assign computation and memory resources to meet diverse performance requirements, ensuring that each application is not negatively impacted by the (potentially erroneous or even malicious) behavior of co-running applications.
In recent years, the community has explored a bevy of techniques for resource management and isolation in MPSoCs. In this talk, I first introduce foundational software-based mechanisms and their limitations. I then discuss how improved hardware support can provide better system observability and diverse isolation guarantees based on application's criticality. I review memory management mechanisms provided by major hardware vendors, as well as proposals by my group and the rest of the community. Finally, I argue for portable abstractions to expose hardware resource management decisions at the hypervisor and OS level.