PhD Course: Privacy Preserving Computation Methods

From July 22nd through 24th, in the meeting room at the first floor of MO27 building (DIEF), Prof. Julián Salas Piñón (https://talent.uoc.edu/es/julian-salas-pinon.html) will lecture the course on "Privacy preserving computation methods: from the basics to Privacy Preserving Machine Learning".

This is a regular course of the PhD school and attendance is warmly recommended.

The course encompasses 12 hours, detailed as follows:

Monday 22:
• Introduction to privacy concepts, cases of attacks to naïve anonymization (2h)
• Pseudonimity, k-anonymity and variants. (1h)
• Activity (1h)

Tuesday 23:
• Differential Privacy (DP) (2h)
• Local DP (1h)
• Activity (1h)

Wednesday 24:
• Privacy preserving techniques for Machine Learning (2h)
• Differentially private supervised and unsupervised learning (1h)
• Activity (1h)

Passing the final exam will award you 3 CFD.

Program:
Data analysis and mining brings many benefits for understanding our society. At the same time, it
reveals a lot about ourselves, from our habits to our personality traits. These characteristics may
also be used to disclose private information, and even to re-identify our data if it is not correctly
protected. To motivate privacy enhancing techniques, we review some of the past attacks to privacy
resulting from naïve anonymization (of medical records, AOL internet search queries, Netflix
movie ratings and NYC-taxis geo-located data) and present measures to prevent attribute and
identity disclosure (namely k-anonymity and differential privacy) in different contexts such as
databases, networks and geo-located data. We finish the course explaining the trade-offs between the risk of disclosure and utility loss in Machine Learning Algorithms for Responsible AI.

We kindly ask to fill the form if you are interested in participating.

https://docs.google.com/forms/d/e/1FAIpQLSeozaYxSWhWy2yX21qsF3pVZKZntcLvtXZ8hpoEnbhvQjF3Qw/viewform?usp=sf_link