S3RI Seminar - "Privacy and Security in Bayesian Inference ", Talk by Dr Louis Aslett (Durham University) Seminar
- Time:
- 14:00 - 15:00
- Date:
- 13 December 2018
- Venue:
- Lecture Theatre 7B, Building 54, Mathematical Sciences, University of Southampton, Highfield Campus, SO17 1BJ
For more information regarding this seminar, please email Dr Helen Ogden at H.E.Ogden@southampton.ac.uk .
Event details
The growth of data sets in Bayesian analyses brings with it concerns surrounding privacy and security, both for the raw data during model fitting and for the potential leaking of sensitive information via the fitted model. This talk will present recent developments combining different privacy and security methodologies to provide protection in the setting of multiple parties wanting to pool their data to produce a Bayesian model fitted on the combined data, considering security and privacy in both the inference and post inference phases.
The seminar will also be available via a live web-cast here
Speaker information
Dr Louis Aslett , Durham University. Louis' current primary research is at the interface between cryptography and statistics, with the focus on privacy preserving statistical analyses. His personal interest is on the statistics side of this fusion, developing novel statistical methodology which is amenable to use in the constrained environment of encrypted computation made possible by recent developments in homomorphic encryption. Louis' other main strand of research is in reliability theory, where interest is in the structural reliability of engineered systems, usually taken from a Bayesian perspective. He also have research interests in computational acceleration of Hidden Markov Models (HMMs) as used in genetics which result in intractable inference as population sizes grow. Threaded through all these research interests is a particular interest in modern massively parallel computing architectures such as GPUs and the development of statistical methodology which is amenable to implementation in such environments.