CSIS Seminar

Cyber security over AI-empowered Internet-of-Things

Abstract

Cyber security is one of the most challenging missions in the era of ubiquitous networks. With the rapid development of the Internet-of-Things (IoT) and various applications of artificial intelligence (AI) in the cyber world, the research in cyber security is expanding beyond the traditional topics focusing on information confidentiality, authentication, and integrity. Many novel vulnerabilities are threatening AI-empowered IoT. My research about cyber security over AI-empowered IoT mainly focuses on two aspects: 1) studying advanced threats that undermine the efficiency of AI in IoT, and 2) proposing solutions to resist these attacks. In this talk, I will introduce several attacks in AI-empowered IoT, including poisoning attacks, adversarial attacks, and credit falsification attacks. Meanwhile, I will discuss how these attacks weaken the system and the challenges to overcome them. After that, I will detailly present one of my works that proposes a secure and efficient federated learning approach from a device selection perspective against the model poisoning attack. Specifically, this work first formulates an optimization problem to minimize training costs while maximizing accuracy. Then, two model poisoning attacks are designed to degrade the performance of training. Finally, two device selection methods are proposed to eliminate the effect of the designed model poisoning attacks. In addition, the main ideas and results of the solutions to the adversarial attack and credit falsification attack also will be briefly presented.

Speaker Bio

Chong Yu is working toward the Ph.D. degree at the Department of Electrical and Computer Engineering, University of Nebraska-Lincoln. She received her B.Sc. degree in communication engineering, M.Sc. degree in communication and information system, and Ph.D. degree in computer system organization from Northeastern University, Shenyang, China. She has 14 publications at prestigious journals and highly selective conferences. Her research interests include cybersecurity, secure and privacy-preserving distributed artificial intelligence (AI), and Internet of Things (IoT).