Security Data Science
The volume of data generated worldwide is increasing rapidly, both on public and private networks. Hence, any malicious behaviour leaves a trail of data behind. This presents Security Data Science with an opportunity to develop and apply techniques from mathematics, statistics and computer science to identify and analyse cyber security and national security threats, using open and closed source data. Security Data Science collaborates with government and private industry to answer questions such as “How can threats existing within private networks be detected?“, “How can robust models of security phenomena be developed?” and “How does disinformation spread through online sources?“. We aim to develop novel techniques, utilising existing data sources, to provide outcomes to minimise exposure of risk to cyber-attacks.
Our current research interests are:
- Graphs and Big Data;
- Ethical thinking in cyber influence, building trust in ML, help counter disruption;
- Information diffusion and propagation in non-traditional social networks;
- Vulnerability modelling and Attack modelling.
Our Research Challenges
Misinformation, polarisation and radicalisation in the online world
How can we identify the trajectory of misinformation and disinformation on social networks, and identify the accounts that are the source? How does polarisation and radicalisation emerge and progress on social networks, and can we find intersecting points of opinion in similar communities? How can we model and trace the evolution of events?
Online Social Networks (OSNs) allow people to easily connect with like-minded others, thus forming communities of interest around intersecting ideologies. This contributes to the formation of echo chambers, whereby people rely on a limited number of “authority” sources of information and quickly reject views from outside their online community. Lack of exposure to information from different perspectives may lead to polarisation of attitudes regarding social, political and religious issues and may foster extremism and radicalisation. Understanding exactly how the (re)organisation and evolution of closed communities/groups in OSNs may be contributing to the formation of echo chambers and potentially polarisation and radicalisation is challenging owing to a lack of appropriate conceptual frameworks and computational tools.
In addition, there are computational and theoretical challenges: the restriction on accessing data from OSNs via APIs, inconsistencies in retrieving data using different tools and lack of tools to measure confidence in the models for making decisions. While traditional models such as machine learning techniques are widely used for predictive modelling on online data, they suffer from multiple challenges. They are likely non-intuitive which makes it difficult for the analysts to understand the logic behind predictions, particularly when there is a plethora of data involved, and the assumption of a self-contained stationary data set that features all events of interest. We are tackling this challenge by incorporating social science theory into our modelling techniques to produce explainable and generalisable models, that are able to provide insight into present and future misinformation campaigns.
Mr Andrew Feutrill
Research Team Leader
P. +61 8 8303 8277
Dr Wei Kang
P. +61 8 8305 0671
Ms Regine Richelle
P. +61 8 8305 0670
Dr Selasi Kwashie
P. +61 8 8305 0693
Dr Geoff Jarrad
Senior Research Engineer
P. +61 8 8303 8869