Security and Privacy in Information Systems: Algorithms and Fundamental Limits

Abstract

Personalized data collection is becoming pervasive, and data is the key enabler that drives applications spanning all sectors of our society, including e-commerce, social networking, and healthcare. On one hand, collecting data at a fine granularity can provide higher utility. On the other hand, without rigorous privacy-preserving mechanisms, there is a risk of potential privacy breaches which often come with a psychological and socio-economic impact. Such privacy breaches are becoming increasingly commonplace, and could be intentional or unintentional. In this dissertation, we consider an information theoretic study of the challenges facing information systems, and propose efficient and practical solutions using recent advances in information theory and machine learning. Further, we focus on addressing four challenges: (1) Study the physical layer security for wireless networks with imperfect channel state information at transmitters, and devising novel secure information transmission schemes under different secrecy constraints. (2) Study a context-aware privacy notion (namely, local information privacy (LIP)) that relaxes the defacto standard privacy notion, i.e., local differential privacy, and lay down the theoretical foundations of LIP and design practical context-aware privacy mechanisms in the local setting. (3) Propose novel communication-efficient transmission schemes for machine learning tasks over wireless networks with provable privacy guarantees. (4) Study the fundamental limits of community detection in social networks, and design novel recovery algorithms while protecting the privacy of individual relationships in the network.