Collaborative Research: CIF: Medium: Do You Trust Me? Practical Approaches and Fundamental Limits for Keyless Authentication

点击翻译按钮获取中文摘要

The last decade has witnessed a wide variety of emerging communication systems that have revolutionized nearly every aspect of society. To support this revolution, further technological advances are required to address ever-growing security and privacy concerns, including combating attempts by adversarial entities at tampering with communications. Despite recent advances in network security, such as public-key cryptography and various security protocols, the current design of security solutions is often governed by the assumption that an adversary is present at all times. In practice, this assumption may waste resources: even though adversaries may appear at any moment, they are not present for a majority of the time. This project explores authentication as an efficient, low-complexity approach to the security of communication systems. Authenticated communication requires transmitting a message from one location to another, so that whenever the receiver accepts a message, it can guarantee the message has not been corrupted. The hallmark of this approach is that it does not require any pre-shared secret knowledge between users, and it incurs little extra cost when an adversary is not present while providing an immediate alarm signal when one is. This research aims to foster interdisciplinary cooperation across engineering and mathematics, and will actively engage both undergraduate and graduate students across institutions.The project addresses information- and coding-theoretic questions related to both fundamental performance limits and principles of optimal authentication code design, along with the application of authentication to emerging areas such as computing and machine learning. First, new fundamental limits for authentication in point-to-point communication systems are being derived, building on the research team’s preliminary work. This includes the study of Gaussian models as well as myopic channels, where the adversary sees a noisy version of the transmitted signal and may use this knowledge to inform its strategy. Next, the project is studying canonical networks by considering a novel objective of partial correction, which allows a receiver to decode a fraction of the transmitted messages, even if it must reject the remainder. These theoretical findings will also be translated into practical coding schemes for authentication. In particular, new code constructions are being developed by incorporating a controlled amount of non-linearity into existing linear coding schemes. Finally, authentication is being interpreted in and applied to the context of emerging applications such as distributed computing and dataset integrity in supervised learning.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.