Machine Learning , Robust Optimisation, and Verification

The scale and diversity of organisational cybersecurity issues are such that it cannot possibly consider every adversary. Decision-makers must consider risk, strategy, service availability and compliance. Led by the Department of Computing, this project will examine new models and approaches to create more robust assessment, monitoring and control of cybersecurity systems in the face of greater uncertainty.

This project takes a multi-Faculty interdisciplinary approach to explore new modelling and analysis capabilities to meet cybersecurity needs. We investigate new approaches to modelling and optimisation that can more robustly assess, monitor, and control cybersecurity systems, processes and infrastructure in the face of uncertainty, with a particular focus on privacy.

The work is divided into four work packages examining differential privacy, detecting, mitigating or preventing cybersecurity threats, robust optimisation for Bayesian networks, and secure multi-party modelling.

Key research tasks:

  • Developing approaches to differential privacy relevant to and applicable in cybersecurity decision-making.
  • Developing robust optimisation techniques to detect, mitigate or prevent specific threats to cybersecurity.
  • Inventing foundations for robust optimisation of causal networks with particular emphasis on Bayesian Belief Networks.
  • Evaluating the research outputs through case studies on systems’ security to “big picture” models of cybersecurity stances, policies, and their consequences.
  • Integrating approaches and lessons learned during the project to offer new capabilities in cybersecurity research, notably Secure Multi-Party Modelling.

Research partners

  • Department of Computing
  • University of Singapore

FunderEngineering and Physical Sciences Research Council (EPSRC)

Duration: March 2016 – December 2018

Business school lead: Dr Wolfram Wiesemann

Principle investigator: Professor Michael Huth (Department of Computing)