Hide-and-Seek Privacy Challenge (NeurIPS 2020)
|Title||Hide-and-Seek Privacy Challenge (NeurIPS 2020)|
|Summary||Building novel methods for privacy-preserving data sharing and/or re-identification|
|Keywords||modelling, privacy preservation, classification, re-identification|
This project is about building competitive methods for NeurIPS 2020 Hide-and-Seek Privacy Challenge: https://www.vanderschaar-lab.com/privacy-challenge/
Competing in this challenge would also be a challenge since the deadline is quite soon (November 15th, 2020), however submitting a reasonable method to this challenge guarantees at least a grade 4 for the thesis.
More realistically, the aim is to take up the same challenge and tackle it throughout your thesis. It provides a great platform with freely accessible dataset and open-source contributions. An evaluation set will be available only to the contributors after the deadline, so it is advantageous to enter the competition officially.
The tasks are: - Hide: Generate synthetical data based on the original dataset so that it is similar to the real data but privacy-preserving (robust to re-identification) - Seek: Classify (re-identify) people accurately from synthetical datasets
The students who are interested in this thesis must have a strong theoretical background in statistics, probability and machine learning and high grades from corresponding courses.