Machine Learning, Robust Optimization, and Verification: Creating Synergistic Capabilities in Cybersecurity Research
 

• Lead PIs : Vincent Tan, Assistant Professor, NUS (vtan@nus.edu.sg ) and Michael Huth, Professor, Imperial College London ( m.huth@imperial.ac.uk )
• Host Institution : School of Computing, NUS
• Partner Institution : Imperial College London

I. Goals

Organizations are faced with an ever growing number of threats on the cybersecurity. Those threats typically involve attempted attacks on the infrastructure, machines and services via integrated malware attack toolkits. A successful hack into the cybersecurity could lead to aftermaths such as the leak of privacy and the loss of high-priced assets. To better address threats of cybersecurity, a model-based decision support for the management of cybersecurity is becoming a practical solution.

The project will investigate new approaches for modelling and optimization by which cybersecurity can be more robustly assessed, monitored, and controlled in the face of uncertainty. Particular attention will be paid to decision support that is both optimal and robust against uncertainty in model parameters, that respects the privacy of individuals, and that protects confidential information.

II. Technologies

This project also brings together research in machine learning, robust optimization, verification and cybersecurity to explore new modelling and analysis capabilities for needs in cybersecurity. We have already made significant progress in distributionally robust optimization with infinitely constrained ambiguity sets, applications of such theories to machine learning problem such as classification, and understanding the fundamental properties of various leakage functions in physical-layer and cybersecurity applications (see for example, arxiv 1709.02168 and 1707.00810). These works have been submitted to top tier journals such as Operations Research and the IEEE Transactions on Information Theory.