Overview
*Bridges to Self* examines how public web scripts and native Android apps used localhost and related channels to connect web identifiers with longer-lived native identity state. The series moves from the mismatch between browser/app isolation and reachable local sockets, through Yandex’s delayed, configuration-driven HTTP(S) bridge and Meta’s WebRTC route for _fbp, to bounded measurements, consent timing, and browser defenses. Its central test follows actors, data, persistence, and back-end use: a local request alone is not proof of tracking. The study reports substantial observed Android deployment and post-disclosure declines, but its counts remain lower bounds—not user totals, universal legal findings, or proof that future variants are impossible.
Learning outcomes
- A localhost endpoint can create a non-local privacy consequence when it joins a web value to native state and an operator back end.
- Meta and Yandex used different transports, but the observed bridge had the same web, local, native, and cloud roles.
- April 2025 Android measurements established substantial observed deployment, while lower-bound limits constrain user, prevalence, and future-variant claims.
- Effective defense must mediate the relationship across protocols and address forms, not only a known domain or port.
Keywords
Preview every chapter · Unlock full text, audio, and video from chapter 4
PlansFAQ
How did Bridges to Self link web and native identities?
Bridges to Self documents a local-channel exchange in which a web identifier reached a native app, was related to an app-side identifier, and was sent to an operator back end. Browser and app storage could remain sandboxed while the identity relationship crossed the interface.
What did the study actually measure?
It measured localhost HTTP(S), WebSocket, and selected WebRTC activity in targeted website crawls and tested popular Android apps, then connected web requests with native listeners and cloud traffic. The counts are conservative lower bounds, not user totals or a census.
Why were Meta and Yandex’s implementations hard to detect?
Meta and Yandex varied transports and used background listeners, delayed activation, remote configuration, obfuscation, or timing limits described in the tested deployments. These conditions could hide the end-to-end chain from browser-only or short app tests.
What defenses protect against localhost identity tracking?
The strongest defense mediates the web-to-native relationship across destinations, address forms, and protocols while making user choice clear and preserving justified integrations. Blocklists and port rules can interrupt known deployments; Local Network Access adds broader permission coverage where implemented, but the paper identifies scope and side-channel limits.