Year: 2022
Type: dissertation
Abstract: Face authentication systems are used more and more in different applications.However, their security relies on the face liveness detection method they use.With the advent of deepfake technologies that make modifications like face swapping easier than ever, there is an urgent need to understand what threats deepfakes pose to liveness detection and explore how to mitigate those threats.In this thesis, I first survey existing methods for face liveness detection.Then I will survey deepfake technologies in terms of its generation and detection methods to provide a comprehensive overview, with a particular focus on the technologies that could be used for practical attacks against liveness detection.I discuss the implications and conduct a case study on a commercial face liveness detection system to show how one can make such systems secure in practice.Finally, I would conclude with open research directions for future work.Foremost, I would like to thank my advisor, Professor Gang Wang, who gives me endless support, encouragement and precious opportunities, and guides me on finding my own research topics.Without the many fruitful discussions with him, this thesis would not come into shape.I would also like to thank all the professors that I have worked with here at University of Illinois.I want to thank Professor Michael Bailey for his course on computer security that sparks my initial interest in this research area, and thank Professor Andrew Miller and Professor Christopher Fletcher for giving me opportunities to explore the different security topics and shaping my way of conducting research.I am very thankful to all my co-authors, Thomas Yurek, Riccardo Paccagnella, Qingying Hao and Jaron Minks.I learned a lot from their research methodology and their research experience.I thank them for always being so kind while answering my questions when I knew little about the topic, helping me get on board and guiding me through every step.I am deeply grateful to my friends, Jingxiang Lin, Zhengqi Fang, Ruian Pan and Yanye Li who are always there for all kinds of discussions that help me relax or get inspired.Additionally, I am incredibly grateful for all the help I receive from many
Author: Luo, Licheng
Institution: University of Illinois Urbana-Champaign
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7220765529
