Feedback and Optimisation for Well-behaved Anonymous Communication Networks

Anonymous communication networks (ACNs), like Tor and mix networks, protect our sensitive communication meta-data, such as whom we talk with, how often we chat and for how long. This meta-data is privacy sensitive since it can be used to reveal secrets that might otherwise be hidden, even when end-to-end encryption is used. This project is timely since mainstream interest in communication privacy on the Internet has grown since the Snowden-revelations about state-level mass surveillance.

However, it is a challenge for ACNs to be deployed since it is hard to tune system parameters that matches the actual realised level of privacy.
This is due to the fact that the privacy, security, and performance of ACNs
is critically impacted by environmental conditions and user behaviour. For example, it is often assumed that
messages are not fragmented---broken up into smaller pieces---as they flow over the network. However,
the reality is that messages are routinely broken up for performance reasons. Similarly, it is assumed that there is a constant level of user activity, however, users tend to have diurnal activity cycles with bursts and lulls throughout the day.

To remedy this current situation, this project aims to bridge the fundamental gaps and provides a framework and a set of methodologies to measure, analyse, and tune ACNs in realistic settings.
It does this by pursuing three objectives:

1. Mapping & Tuning: New analysis to uncover and formalise relationships between abstract security parameters and real-world network measurements with the view to optimally tune the ACN.
2. Feedback: Investigate novel ACN designs with feedback loops that provide the ability to automatically tune security parameters at run time.
3. Use-case validation: Evaluate in targeted use-cases of email, web-browsing, and IoT data collection systems to validate the automated tuning methodology.

A common occurrence motivates the need for this project. Let us consider an email provider desiring to provide user anonymity as a market differentiator. Referring to the state-of-the-art in the email-securing mix networks literature it is difficult for the non-expert to reason how to correctly parametrise the mix network for the email provider's particular user base. Mapping & Tuning are the missing ingredients holding back deployment. Given a tuned ACN at start-up time, Feedback can be employed to automatically set and adjust the security parameters necessary for the email anonymity service at run time. The provider nor its system administrator needs to become expert in ACN design nor the abstract privacy metrics necessary for manual tuning.

This project will leverage recent advancements in the design of mix networks and privacy-preserving network data collection as the basis of our building blocks from which we can extend and enhance. The lasting positive impact of the resultant trustworthy intelligent and adaptive ACNs will be increased adoption and therefore robust privacy for the UK and global public. The technology, data-sets, and tooling developed will open-sourced and will be a boost to the UK privacy technologies marketplace.