trackers — Web domains and trackers (2012)

Description

Bipartite networks of web domains and the third-party services embedded in their pages, from the large-scale analysis of online tracking by S. Schelter and J. Kunegis. The data were extracted from the 3.5 billion web pages of the CommonCrawl 2012 corpus, a breadth-first crawl of the web: for each page, the extractor collected the pay-level domains (the sub-domains of a public top-level domain that are usually registered by a paying owner, e.g. facebook.com) of the external resources embedded in its HTML code, i.e. those referenced in the src attribute of script, iframe, link and img tags, and those found as URI strings in its embedded JavaScript code (third parties loaded dynamically by external JavaScript code cannot be found). The data were aggregated by pay-level domain, so that both web sites and third parties are represented by their pay-level domains. The 1,375 third parties found in at least 0.01% of the domains were labeled by hand according to their purpose and business model (e.g. advertising, analytics, widgets, content delivery, hosting), and 355 of them were deemed to be potentially privacy-threatening tracking services, whose purpose or business model aims at recording the behavior of the users visiting the pages that embed them (advertising, analytics, widgets and beacons).

The collection comprises two networks:

third_parties: the third-party network, obtained from the Koblenz Network Collection (KONECT), with 27,665,730 domains and 12,756,244 third-party domains, and 140,613,762 edges. An edge connects a domain to a third party embedded in its pages. Each domain is represented by different nodes as a web site and as a third party. The nodes are not identified by their domain names, and have only the property is_right, which is true for the third parties.
trackers: the tracking network, from the web page of the authors, with the domains that embed at least one of the 355 tracking services, and the trackers themselves. An edge connects a domain to a tracker embedded in its pages. A domain that is also a tracker (e.g. facebook.com) is represented by different nodes as a web site and as a tracker.

The trackers network has the following properties:

(Node) name: the pay-level domain.
(Node) pld_id: the ID of the domain in the pay-level domain index of the source data.
(Node) is_tracker: whether the node is a tracker (otherwise it is a web site).
(Node) category: the category of the tracker (Advertising, Analytics, Widget or Beacon; empty for web sites).
(Node) company: the company owning the tracker, as determined by hand by the authors (empty for web sites).
(Node) registration_country: the country of registration of the domain of the tracker (two-letter code, # if unknown; empty for web sites).
(Node) embeddings_javascript, embeddings_iframe, embeddings_image, embeddings_link: the number of domains that embed the tracker as JavaScript code, in an iframe, as an image, or as another linked resource (e.g. style sheet), respectively (0 for web sites). The total number of domains that embed the tracker, also given in the source data, is the degree of the node.

The source data also give the organization that registered the domain of each tracker, which was left out as it mostly coincides with the owning company.

Tags

Informational Web graph Unweighted Metadata

Added

2020-08-11

Citation

S. Schelter, J. Kunegis, "Tracking the Trackers: A Large-Scale Analysis of Embedded Web Trackers." Proceedings of the International AAAI Conference on Web and Social Media 10(1), 679-682 (2016) 10.1609/icwsm.v10i1.14769 Sci-Hub
S. Schelter, J. Kunegis, "On the Ubiquity of Web Tracking: Insights from a Billion-Page Web Crawl." The Journal of Web Science 4(4), 53-66 (2018) 10.1561/106.00000014 Sci-Hub

Upstream URL OK

https://ssc.io/trackingthetrackers/
http://konect.cc/networks/trackers-trackers

Upstream license

Networks

Tip: click on a column header to sort the list, or hover over it for a description.

Column legend

$\left<k\right>$: Mean degree
$\sigma_k$: Standard deviation of the degrees
$\lambda_h$: Largest eigenvalue of the Hashimoto (non-backtracking) matrix
$\tau$: Random walk mixing time
$r$: Degree assortativity
$c$: Global clustering coefficient
$\oslash$: (Pseudo-) diameter
$S$: Relative size of the largest component
NPs: Node properties
EPs: Edge properties

Name	Nodes	Edges	$\left<k\right>$	$\sigma_k$	$\lambda_h$	$\tau$	$r$	$c$	$\oslash$	$S$	Kind	Mode	NPs	EPs	gt	GraphML	GML	csv
third_parties	40,421,974	140,613,762	6.96	2294.88	2673.57	13778.00	-0.10	0.00	21	0.99	Undirected	Bipartite	is_tracker		270.5 MiB	625.5 MiB	601.7 MiB	625.0 MiB
trackers	18,783,254	36,982,655	3.94	2987.05	1969.09	76.09	-0.30	0.00	6	1.00	Undirected	Bipartite	is_tracker name pld_id category company registration_country embeddings_javascript embeddings_iframe embeddings_image embeddings_link		137.4 MiB	290.5 MiB	294.3 MiB	306.1 MiB