Diffusion of User Tracking Data in the Online Advertising Ecosystem

Abstract

Advertising and Analytics (A&A) companies have started collaborating more closely with one another due to the shift in the online advertising industry towards Real Time Bidding (RTB). One natural way to understand how user tracking data moves through this interconnected advertising ecosystem is by modeling it as a graph. In this paper, we introduce a novel graph representation, called an Inclusion graph, to model the impact of RTB on the diffusion of user tracking data in the advertising ecosystem. Through simulations on the Inclusion graph, we provide upper and lower estimates on the tracking information observed by A&A companies. We find that 52 A&A companies observe at least 91% of an average user’s browsing history under reasonable assumptions about information sharing within RTB auctions. We also evaluate the effectiveness of blocking strategies (e.g., AdBlock Plus), and find that major A&A companies still observe 40–90% of user impressions, depending on the blocking strategy.

Editors’ Introduction

“…Along the first direction, privacy enhancing tools such as Mozilla Lightbeam [20], uBlock Origin [56], and Ghostery [22] provide users transparency and control over online tracking. Our work can be used to address known limitations [36] of these tools by identifying server-side data sharing practices of online trackers. Along the second direction, our HB measurements can be used to improve existing user valuation tools such as RTBAnalyzer [67] and YourAD-Value [71] by capturing a more complete picture of all advertisers' bidding behaviors.…”

“…Since the user's browser is not involved in the mechanism, these data sharing relationships are not directly observable at the client. Instead, more complex controlled experiments are needed to infer such relationships [35,36].…”

“…The underlying intuition was that if a retargeted ad was served by an entity that did not observe the original visit which triggered the retargeted ad, then it must have got information about this visit through an S2S information flow. We leverage the same underlying intuition as in Bashir et al [35,36] to infer tracker-advertiser relationships. However, instead of relying on retargeting as the "signal" for data exchange, we operationalize this intuition using a machine learning model that is trained to pre-dict an advertiser's bid values using presence/absence information of popular trackers.…”

Quantity vs. Quality: Evaluating User Interest Profiles Using Ad Preference Managers

“…Participants were recruited in two waves, with the first wave paid $5 for completing our survey, and the second paid $4. 9 This data was collected in March and May 2018.…”

“…Despite widespread concerns about online privacy and attention on the online advertising industry, there is one specific aspect of this ecosystem that remains opaque: what information is actually contained in advertising interest profiles, and are these profiles accurate? The extant literature presents a very complete picture of who is tracking users' online [7], [27], [9], [86], [71], [70], as well as how tracking is implemented (e.g., fingerprinting) [37], [78], [5], [57], [60], [72], [58], [1], [40], [28], but not necessarily the what. Controlled studies have shown that advertising platforms like Google do indeed draw inferences about users from tracking data [89], [20], [46], [47], but this does not address the broader question of what platforms actually know about users in practice.…”