Abstract

Abstract:

Biometric information, such as facial and vocal features are critical for identifying and authenticating individuals. This is an enabling service for smart-spaces, allowing building management agents to easily monitor `who is where’, anticipating user needs and tailoring the local environment and user experiences. Although biometric recognition, especially through the use of deep neural networks, has achieved stellar performance with large datasets, the majority of approaches require supervised learning, that is, to be trained with tens or hundreds of images of users in different conditions. In the first part of my talk, I will talk our recent work on using cross-modality observations from wireless signals to supervise different biometric recognition systems. By learning and refining the noisy and weak association between a user’s smart-phone and her biometric observations, we show that a deep neural network can be fine-tuned and tailored to the environment, users and conditions of a particular space, e.g., lighting variability and viewing angles. Following the same thread, the privacy and security implications of cross-modality inference will be discussed in the second part of my talk. I will introduce the leakage of passwords, for both PINs and android pattern locks, keystroked on the touch screens. Particularly, I will talk how the trained attack model with limited passwords can generalize to infer the whole universe of passwords.

Bio:

Chris Xiaoxuan Lu is a Ph.D student in Computer Science at Oxford University, co-advised by Professor Niki Trigoni and Andrew Markham. He received his M.Eng degree from Nanyang Technology University, Singapore. He is broadly interested in ubiquitous and mobile computing, with a focus on enabling ambient intelligence and calm technologies with cross-modality solutions.