Knowledge Representation and Information Security

点击翻译按钮获取中文摘要

This research program will focus on using tools and methods from the formal Artificial Intelligence(AI) community to solve problems in Information Security. We will be particularly interested in using models of reasoning to identify and mitigate vulnerabilities related to three key issues: - Trust: When can we believe information received from another agent, or allow them to act on our behalf? - Deception: How can we identify when an agent is actively trying to manipulate our beliefs? - Privacy: How can we ensure that private data is not divulged inadvertently in communication? These issues will be addressed by developing explicit, declarative models of the knowledge and beliefs of communicating agents. We will use our models to develop tools that can solve practical problems of interest in the Information Security sector. This program falls under the umbrella of Knowledge Representation and Reasoning.  Fundamentally, our approach is to view secure communication as a reasoning problem: agents exchange messages and draw conclusions from information.  Modelling this kind of reasoning requires a clear understanding of trust, and how beliefs change in response to new information.  We have several distinct goals.  One goal is theoretical: we would like to extend existing models of knowledge and belief to capture key concepts required for reasoning about secure communication.  The second goal is practical: we would like to implement software tools that can efficiently solve reasoning problems in the formal models that we have developed. The third goal is applied: we would like to use our tools to reason about difficult issues in Information Security. The problems of interest will be motivated by interaction with industry partners, with an emphasis on practical problems that involve reasoning about trust and belief. This research program is novel because it aims to move formal models of belief from the idealized setting of Logic to the applied context of Information Security. This is significant because it connects two distinct research communities.  One one hand, we have the formal AI research community, which has produced over four decades of rigorous formal research on the dynamics of belief.  While this work has advanced theoretically, it has not been applied extensively at a practical level.  On the other hand, we have the Information Security community.  This is simply a research area that offers an abundance of complex problems that are of great practical interest.  Connecting these two groups therefore makes sense on both sides, providing important practical applications for formal AI while offering the hope of novel solutions to hard problems in the Security community.