Measure The Web

User Tools

Site Tools

Trace:
privacy:email_tracking

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
privacy:email_tracking [2026/09/02 07:40] – Generic + re-verification review fixes: state the real hand-map accounting (183 mapped / 125 with a written reason / 171 defaulted at title level) instead of claiming 183 hand-written reasons; correct '73 papers outside the pool' to 67 distinct / 72 menti karel.kubicek.claudeprivacy:email_tracking [2026/10/08 12:19] (current) – Link and list privacy:pii_leakage (new page). Authored by Claude. karel.kubicek.claude
Line 107: Line 107:
 The other half of the topic, and the half that is growing. An email address or phone number is a **cross-context identifier that survives everything third-party cookies do not**: it is stable, the user hands it over voluntarily, and it can be hashed and matched into an advertising platform. The other half of the topic, and the half that is growing. An email address or phone number is a **cross-context identifier that survives everything third-party cookies do not**: it is stable, the user hands it over voluntarily, and it can be hashed and matched into an advertising platform.
  
-  * **How the address escapes the page.** Starov et al. {[starov2016_sure]} measured contact forms as the point where a pseudonymous visitor becomes identified. Six years later Senol et al. {[senol2022_leaky]} measured leakage **before submit**: the address was exfiltrated on **1,844** EU and **2,950** US websites, and 41 of the receiving domains were on none of four blocklists. Acar et al. {[acar2020_noboundaries]} found the same behaviour in session-replay and login-manager scripts, and Kieserman et al. {[kieserman2025_tracker]} showed that whether it happens depends on how a given tracker installation is //configured//, not on which tracker it is.+  * **How the address escapes the page.** Starov et al. {[starov2016_sure]} measured contact forms as the point where a pseudonymous visitor becomes identified. Six years later Senol et al. {[senol2022_leaky]} measured leakage **before submit**: the address was exfiltrated on **1,844** EU and **2,950** US websites, and 41 of the receiving domains were on none of four blocklists. Acar et al. {[acar2020_noboundaries]} found the same behaviour in session-replay and login-manager scripts, and Kieserman et al. {[kieserman2025_tracker]} showed that whether it happens depends on how a given tracker installation is //configured//, not on which tracker it is. The measurement recipe behind all four — canaries, the interaction that puts a value on the page, the encoding search and its blind spots — is on [[Privacy:PII leakage]], which owns personal-data leakage in general; this section keeps the address as an identifier.
   * **What the address is worth to an ad platform.** Venkatadri et al. measured this twice: {[venkatadri2018_privacy]} showed PII-based targeting could be inverted — 18 of 20 volunteers who visited a page were correctly identified from a tracking-pixel audience, and phone numbers were inferred from addresses — and {[venkatadri2019_investigating]} tracked where the platform's PII comes from, including that a phone number added purely for two-factor authentication became advertising-targetable after **22 days**. {[venkatadri2019_auditing]} then measured the offline-broker side: over 90% of targetable US Facebook identities carried at least one broker attribute, and coverage was strongly unequal — 64.3% of US counties reached 80% coverage against 13.8% of the poorest decile.   * **What the address is worth to an ad platform.** Venkatadri et al. measured this twice: {[venkatadri2018_privacy]} showed PII-based targeting could be inverted — 18 of 20 volunteers who visited a page were correctly identified from a tracking-pixel audience, and phone numbers were inferred from addresses — and {[venkatadri2019_investigating]} tracked where the platform's PII comes from, including that a phone number added purely for two-factor authentication became advertising-targetable after **22 days**. {[venkatadri2019_auditing]} then measured the offline-broker side: over 90% of targetable US Facebook identities carried at least one broker attribute, and coverage was strongly unequal — 64.3% of US counties reached 80% coverage against 13.8% of the poorest decile.
   * **The address is not one identifier.** Wu et al. {[wu2026_email]} measured alias handling — dots, plus-addressing, sub-addressing — and found providers disagree about whether two strings are the same account. If your method matches on the address, this is your false-positive and false-negative source, and it is provider-dependent.   * **The address is not one identifier.** Wu et al. {[wu2026_email]} measured alias handling — dots, plus-addressing, sub-addressing — and found providers disagree about whether two strings are the same account. If your method matches on the address, this is your false-positive and false-negative source, and it is provider-dependent.
Line 336: Line 336:
  
   * [[Programming:Registration]] — getting an account, which is upstream of getting mail. This page is what the mailbox then lets you measure.   * [[Programming:Registration]] — getting an account, which is upstream of getting mail. This page is what the mailbox then lets you measure.
 +  * [[Privacy:PII leakage]] — personal data in general (names, phones, search terms, health and wallet data, the address included) reaching third parties from a page: the canary, the encoding search, configuration versus transmission.
   * [[Privacy:Requests]] — the web-page pixel, filter lists, and request classification. The other "pixel", with a different method.   * [[Privacy:Requests]] — the web-page pixel, filter lists, and request classification. The other "pixel", with a different method.
   * [[Privacy:Consent]] — the banner in front of the registration form you are about to fill in.   * [[Privacy:Consent]] — the banner in front of the registration form you are about to fill in.
privacy/email_tracking.txt · Last modified: by karel.kubicek.claude