PhD in Computer Science

Privacy for Linked Data

Academic year 2012/2013

Sommario del corso

• Program
• Notes
• Teaching tools

Program

Web of Linked Data aims at the introduction of common format and principles for publishing data on the Web, such that they can be easily interconnected and exploited. The current recommendation is to use URIs as names and RDF directed graph data, which can be consumed by use of a query language such as SPARQL. The lack of privacy mechanisms often discourages people from publishing data in the Web of Linked Data. We propose a formal model which ensures privacy protection of a single RDF triple. In our calculus, processes are those from calculus for Linked Data of Horne and Sassone, associated with parallel composition of RDF triples. These triples describe the web-profile of the user that aims at performing an action. An administrator of a data graph in the observed system defines a privacy protection policy for each triple of the data in RDF format. A process running on behalf of a user is allowed to access a triple if the privacy protection policy gives the positive answer applied to the user's profile. The introduced type system verifies policy conformance, assuring that in a well-typed network:

- the privacy protection policy of each stored triple agrees with the privacy administration policy of the graph it belongs to. This induces the property that if a process has a privilege to administrate policies in a data graph, it can access each triple in the graph,

- a process can administrate the privacy protection policy of a stored triple in a data graph only if its profile contains a property that is in the administration policy of the data graph and

- a process can read a stored triple only if its profile contains a property that is in the privacy policy of the stored triple.

Class schedule

Note

SEMINAR HELD BY:

Svetlana Jakšić

(http://imft.ftn.uns.ac.rs/~svetlana/)

University of Novi Sad

• Teaching materials
• Exams and finals
• Go to Moodle
• Visit the forums