Web Transparency AT Princeton

WEB TRANSPARENCYAT PRINCETONPRESENTER: Christian EubankPAPER: Shining the Floodlights on Mobile Web Tracking [E., Melara, Perez-Botero, Narayanan]

Original Motivation

•Mobile tracking knowledge vacuum

•Mobile devices are everywhere

•Perceived measurement difficulty

Guiding Questions•How do companies track mobile devices differently?

•What about tablets?

•Physical v. emulated devices?

Our Approach

•Data collection via FourthParty

•FP Database Schema

•Mixture of manual/automated analysis

Mobile JavaScript crawler

•Originally considered:

•Future: desktop crawls with Selenium

Findings: Few Mobile-Only Networks

•

•

•Otherwise, .mobile and m. addresses

Findings: Devices Roughly SimilarDesktop Asus Galaxy Phone E. Tab E. Phone

Desktop 1.00 0.77 0.82 0.76 0.75 0.75Asus 1.00 0.87 0.83 0.84 0.83Galaxy 1.00 0.88 0.88 0.87Phone 1.00 0.86 0.95E. Tab 1.00 0.86E. Phone 1.00

Findings: Cookie Longevity

Desktop Asus Galaxy Phone E. Tab E.Phone0

1

2

3

4

5

6

7

First PartyThird Party

Device

Coo

kie

Expi

ratio

n Ti

me

(day

s)

Growing Cookies: yieldmanager’s bhCookie Value String DomainEXAMPLE 1"b!!!!#!!2-]!!!!#>+YEL" Netflix"b!!!!$!!2-]!!!!#>+YEL!%HWu!!!!#>+YE]" Cracked"b!!!!%!!2-]!!!!#>+YEL!%HWu!!!!#>+YE]!%ODP!!!!#>+YF$" SalonEXAMPLE 2"b!!!!#!%HWu!!!!#>+YG<" Cracked"b!!!!$!%HWu!!!!#>+YG<!%ODP!!!!#>+YGC" Salon"b!!!!%!!2-]!!!!#>+YGP!%HWu!!!!#>+YG<!%ODP!!!!#>+YGC" NetflixEXAMPLE 3"b!!!!#!%ODP!!!!#>+YLs" Salon"b!!!!$!!2-]!!!!#>+YM$!%ODP!!!!#>+YLs" Netflix"b!!!!%!!2-]!!!!#>+YM$!%HWu!!!!#>+YM,!%ODP!!!!#>+YLs" CrackedBold text indicates site ID; Underlined text indicates user IDHwu = cracked.com; 2-] = netflix.com; ODP = salon.com

Growing Cookies: Takeaways

•Third parties storing on client

•Placed by 8% of desktop third parties

•Possible growing cookie library

Next Steps

•Continued Cookie Analysis

•Deeper Crawls

•Exploring rich data from FourthParty • Our data open to other researchers

Automated Framework

WWW

Controller DATA

Overarching Goals•Keep public informed

•Help to shape policy

•Promote web transparency

• Individual case studies under umbrella of web transparency

Leveraging our infrastructure

•Automatic social media interaction

•Detecting cookie synchronization

Conclusion

•Need for web transparency

•Currently building measurement tool

•We invite collaboration

QUESTIONS?