Abstract: Smart devices are becoming increasingly affordable and ubiquitous; whether they are seamlessly embedded in the environment, or carried by people and robots, they continually generate sensor data about the world that surrounds them. Recent advances in machine learning have revolutionised the ability of smart sensors to perceive and interpret context, infer human activities and react to human preferences. This talk will highlight challenges and opportunities in designing machine learning techniques to solve problems of context inference and control. I will then show how the power of machine learning, can become a severe vulnerability, in the absence of sufficient measures to protect the privacy of individuals.
Bio: Niki Trigoni is Professor at the Oxford Department of Computer Science, heading the Cyber Physical Systems Group. Her interests lie in the tight integration of sensing and machine intelligence for context inference, control and human-machine interaction using a variety of sensor modalities, including inertial, visual, magnetic and radio signals. She has applied her work to a number of application scenarios, including asset monitoring for construction sites, mobile autonomy with humans and robots, and track worker localisation for safety and efficiency. She is also the Director of the Centre for Doctoral Training on Autonomous and Intelligent Machines and Systems in Oxford and Founder of the Navenio Oxford spinout.