New Techniques for Finding and Analysing Information Leaks

We all rely on the security of computer systems in our daily lives; from using a credit card or ATM, to checking our e-mail, we need the computer systems around us to keep our secrets and to preserve our privacy. It would be misleading to think of these systems as either perfectly secure, or entirely broken and open to abuse. For example, some systems can be broken by brute force but still provide some protection against a casual attacker, whereas other systems might leak a small amount of information that could be exploited by an attacker over time. Understanding and measuring the different levels of security that a system might offer is vital if we are going to develop a safe, efficient digital world.

This project will develop new, effective techniques to find and measure security flaws in computer systems. We will use information theory to measure how much information an attacker can learn about the secret information inside a system by observing its public outputs. The key novelty of our approach is to use a combination of statistics and information theory to measure how secure a system is from trial runs of that system.

We will use concepts, such as network information theory and differential entropy, to develop general definitions of quantitative security. These concepts have never before been applied in the field of computer security, and they will lead to better, more expressive definitions that can be applied in a wide range of situations.

Using our theoretical work, we will develop automatic analysis tools, and we will use these to assess the security of a range of systems, including the Freenet anonymity system and RFID tags. We hope that this will serve as an example to other developers and researchers, showing that our tools make information theory-based analysis methods practical and easy to use. This may potentially lead to improved security of many commercial computer systems and faster, easier ways to find information leaks.

This project will develop new, effective techniques to find and measure security flaws in computer systems. The theoretical results we develop will benefit researchers working in the areas of information leakage and information theory. The project we propose combines elements of the formal analysis of information flow and the statistical estimation of information theory. Both of these communities could benefit from this project. Applied security researchers, who currently look at graphs of data or use simple statistical tests to look for information leaks, will benefit from the tools developed in this project, as will more theoretical researchers, who will benefit from ways that their work could be applied to implemented systems

We will increase the impact of our work for security practitioners by building tools that use our methods to automatically analyse data sets. These tools will be based on the rigorous mathematical methods we develop, but will not require any specialist knowledge to use, giving them the widest possible audience. We will make these tools publicly available and promote them by giving talks and demonstrations at seminars and at tool-based conferences.

We will apply the methods and tools we develop to a range of systems, in order to analyse and improve the security they offer. This will serve as an example to other developers and researchers, showing that our tools make information theory-based analysis methods practical and easy to use. This may potentially lead to improved security of many commercial computer systems and faster, easier ways to find information leaks.

The social impact of previous work on e-passports by the principal investigator, and iTunes and online banking security by other group members, is illustrated by the media coverage it has received in highly influential on-line news sites, such as "The Register" and in the national British press, such as "The Daily Mail" and "The Times". This media coverage empowers consumers to make informed decisions about the online services and electronic products they use.

The systems we will examine include the Freenet anonymity network and RFID bank cards. Freenet is a censorship-resistant peer to peer network that aims to protect freedom of speech, particularly in undemocratic, authoritarian countries. Our work with the Freenet project will enhance the freedom of speech of people that use this network to resist censorship by oppressive regimes. Our proposed work on RFID bank cards may turn up problems that can be fixed, so improving the security of the cards. Alternatively, our work may determine that the cards are secure, so increasing the level of acceptance and uptake of this new technology.