Tracking Classification: A Cluster Analysis Approach to Identify Missing Values (Presentations Slides)

Abstract

Presentation part of the Dagstuhl seminar 'Online Privacy and Web Transparency.' This is motivated by the changing legal framework in the European Union, i.e., the General Data Protection Regulation and the (draft) ePrivay Regulation. Transparency and fairness are key elements when it comes to informed consent. The problem with real-time bidding however is that many third-parties are unknown at the time of asking consent from a user. The talk will explore a methodology for classification of third-parties by looking at information flows between real-time bidding networks using cluster analysis. Interactive HTML-widgets will be used to guide the exploration. The difference in effectiveness of ad-blockers in the EU versus the US will also be discussed.

The presentation leads to two conclusions against the background of regulatory changes in the EU (GDPR and ePrivacy Regulation). Conclusion on technology: with cluster analysis on a standard referer graph it is possible to differentiate between actors and their role in interconnected RTB systems. The following cluster analysis approaches have proven to be useful: (a) node betweeness (b) cluster-edge betweenness (c) eigenvector centrality. Conclusion on policy: cookie enforcement can be effective as the example in Slovenia shows.

Keywords: web tracking, big data, privacy, data protection