Thread by @WolfieChristl

2017 conference paper on cross-device tracking based on mass surveillance data from comScore, covering the majority of the US population.

They show how they identify which browsers, smartphones and other devices belong to which individuals and families: pages.cs.wisc.edu/~pb/kdd17_fina…

The paper was written by comScore researchers.

They used personal data on website visits and app/device usage recorded from unaware users over 6 weeks, including 1.2bn Google/Apple 'Advertising' IDs and pseudonymous IDs stored in cookies, in combination with IPs and timestamps.

In another paper published by a similar team in 2018, also based on comScore data, they additionally included device/browser fingerprints (based on e.g. screen size, user agent) and behavioral fingerprints (apps used and domains of the websites visited): pages.cs.wisc.edu/~pb/kdd18b_fin…

Both papers are interesting. Because of their findings (although some kind of quality check may be required), but above all, both make robust claims about comScore's personal data processing.

For example, the 2018 paper reveals comScore is even collecting data from routers (!)

"ComScore Networks' Marketscore application is installed on more than 1 million PCs in the United States …'We do capture information, including data that occurs in secure sessions, to get information like what a person buys'"

CNET Article from 2004 (!): cnet.com/news/comscore-…

"Marketscore is a downloadable application that purports to speed up Internet surfing, and in partnership with Symantec, protect e-mail from viruses"

This is how comScore has been collecting 'clickstream' data on website visits in 2004.

…reminds me of something /cc @josephfcox