Challenges in Learning from Behaviour

Welcome to AFSecurity at UiO,

David Aspinall and Robert Flood from the University of Edinburgh present a talk about challenges for building robust ML models.

See the program below and at the AFSecurity wiki.

TIME: 9 June 2023, at 14:00h
PLACE: Seminar Room Caml (3438), 3rd floor, IFI, UiO.

All interested are welcome. Coffee and snacks served.

AGENDA:
14:00h Welcome to AFSecurity at UiO
14:15h Invited talk

• TITLE: Challenges in Learning from Behaviour
• SPEAKERS: David Aspinall and Robert Flood, University of Edinburgh
• ABSTRACT: Artificial Intelligence and Machine Learning are promoted as solutions for cyber security which can adapt to changes in attacker behaviour, learning models from data to distinguish what is normal and what may consitute a threat. Despite much research, there are still fundamental challenges in each stage of the process which prevent us having good benchmarks to understand the robustness of state-of-the-art models. The challenges span: collecting and sharing data safely, ensuring a sufficient fidelity and diversity of data, and then understanding model performance. We will consider these challenges in turn and present some recent research results which address each aspect. Under the second challenge, we will introduce some work-in-progress on identifying "bad smells" for cyber security datasets used for training.

15:00h Discussion.

BIO: David Aspinall is Professor of Software Safety and Security at the University of Edinburgh. He is Director of the University's UK government recognised Academic Centre of Excellence in Cyber Security Research and was a Fellow of the Alan Turing Institute from 2015-2021. His research interests range from foundations for formal methods in software security to more applied areas including anomaly detection in network security.

Robert Flood is a PhD student in the University of Edinburgh where he previously studied for an MSc in Computer Science. His PhD primarily focuses on synthetic data generation for training anomaly detection systems. He has also worked with cyber security data anonymisation and software correctness with The Turing Institute and The University of Edinburgh.

Academic Forum on Security is organised by