DNS weighted footprints for web browsing analytics. (1st June 2018)

• Record Type: Journal Article
• Authors: García-Dorado, José Luis; Ramos, Javier; Rodríguez, Miguel; Aracil, Javier
• Is Part Of: Journal of network and computer applications. Volume 111(2018)
• Page Start: 35
• Page End: 48
• Publication Date: 2018-06-01
• Subjects: Browsing analytics -- Data monetization -- DNS footprint -- Web-user profiles -- DNSprints; Microcomputers -- Periodicals; Computer networks -- Periodicals; Application software -- Periodicals; Micro-ordinateurs -- Périodiques; Réseaux d'ordinateurs -- Périodiques; Logiciels d'application -- Périodiques; Application software; Computer networks; Microcomputers; Periodicals; 004.05; 004
• DOI: 10.1016/j.jnca.2018.03.008
• Languages: English
• ISSNs: 1084-8045

Abstract

Abstract: The monetization of the large amount of data that ISPs have of their users is still in early stages. Specifically, the knowledge of the websites that specific users or aggregates of users visit opens new opportunities of business, after the convenient sanitization. However, the construction of accurate DNS-based web-user profiles on large networks is a challenge not only because the requirements that capturing traffic entails, but also given the use of DNS caches, the proliferation of botnets and the complexity of current websites (i.e., when a user visit a website a set of self-triggered DNS queries for banners, from both same company and third parties services, as well for some preloaded and prefetching contents are in place). In this way, we propose to count the intentional visits users make to websites by means of DNS weighted footprints. Such novel approach consists of considering that a website was actively visited if an empirical-estimated fraction of the DNS queries of both the own website and the set of self-triggered websites are found. This approach has been coded in a final system named DNSprints . After its parameterization (i.e., balancing the importance of a website in a footprint with respect to the total set of footprints), we have measured that our proposal is able to identify visits and their durations with false and true positives rates between 2 and 9% and over 90%, respectively, at throughputs between 800, 000 and 1.4 million DNS packets per second in diverse scenarios, thus proving both its refinement and applicability.