Title: Probabilistic approaches to the protection of confidential information
Lecturer: Catuscia Palamidessi
Period: April 14,15 Seminari Est, 15-17; April16 Seminari Est 9-11; April 17,18 Seminari Est 11-13
We will present the information-theoretic approach to information flow, focussing on the cases of Shannon entropy and of Rényi min-entropy.
We will link these notions with the concept of adversary.
Then we will present a recent more general approach, based on decision theory.
In the second part of this course, we will introduce the framework of differential privacy, we will present the characterizations and properties of this notion, with emphasis on its compositionality and its independence from the prior information of the adversary.We will also highlight the relation with quantitative information flow.
We will present implementations of differential privacy, and discuss the trade-off between privacy and utility of the data.
Finally, we will present an extension of differential privacy, and its application to privacy-friendly geolocation.