Detecting Anomalies (and Fraud) in Networks:

TARC Workshop: 24th June 2020  13.00-17.30 BST

Network Analysis, Machine Learning and Statistics: How can Policy-Institutes utilize networks to achieve aims such as fraud detection?

There is an explosion of data obtained from systems that can be conceptualized as networks. The collection and analysis of network data plays a key role in a wide range of scientific fields. Examples include, but are not limited to, biology, computer science, sociology and economics.

The analysis of the observed networks requires the ability to examine big data by utilizing advanced data analytical methods. 

  • How can recently developed Machine Learning and Statistical techniques assist organisations (such as Revenue Authorities) to enhance their understanding of the observed networks?
  • Can advanced data analytics improve the detection of anomalies (fraud) in large and complex networks such as Values added Tax (VAT) and Cyber-Security systems?
  • Which are the Mathematical and Statistical challenges?
  • Which are the most relative tools from the emerging literature that can be used?

 These are some of the questions the TARC workshop aimed to provide answers to. Click on the links to download the presentations.

Workshop programme

13:00-13:30 Niall Adams, Imperial College London, UK

Anomaly Detection in Enterprise Cyber-Security 

13:40 - 14:10 Christos Kotsogiannis and Petros Dellaportas, University of Exeter / University College London, UK

Detecting network anomalies in the Value Added Taxes (VAT) system 

14:20-14:50 Michalis Vazirgiannis, LIX Ecole Polytechnique, France

Graph Mining for Fraud detection

15:30-16:00 Simone Gabbriellini, University of Trento, Italy

Network effects on tax compliance 

16:10-16:40 Theodoros Rapanos, Södertörn University, Sweden

Imperfect information, social norms, and beliefs in networks 

16:50-17:20 Sofia Olhede, The École polytechnique fédérale de Lausanne (EPFL), Switzerland

Modeling Networks and Network Populations 

17.30 Workshop close