Security and Privacy in Machine Learning and CPS
GOAL: Study on the security and privacy of machine learning models and other applications in IoT and cyber-physical systems.
Mobile devices and cyber-physical systems (CPS) are equipped with numerous sensors which allow them to offer efficient and effective personalized services and applications. For example, connected and autonomous vehicles (CAVs) feature advanced sensing capabilities, including multiples of range sensors (Lidar and Radar), 360° cameras, onboard GPUs, and high-speed connectivity: Tesla Motors uses a forward radar, a front-facing camera, and multiple ultrasonic sensors to enable its Autopilot feature; Google’s and Apple’s version of CAV uses Lidar and cameras to support autonomous driving; Ford and Uber are also actively experimenting with CAVs.
These advanced capabilities open up a plethora of exciting opportunities for next generation services related to better localization and navigation and traffic optimization. At the same time, their reliance on sensing data and machine learning algorithms for route prediction, collision avoidance and object detection and recognitions, introduces new attack surfaces. Given the widening gap between autonomy and security in this application domain, in tandem with their safety repercussions, there is an impending need for novel solutions that can guarantee trusted outcomes from such sensor-fusion and machine learning algorithms.