Privacy/Roadmap/2014: Difference between revisions

Privacy/Roadmap/2014 (view source)

999 bytes removed , 25 March 2014

Confirmed users

238

edits

@@ Line 16: / Line 16: @@
 == Goal: Firefox users know when they are being tracked ==
 * (Research, implementation) Use machine learning to classify tracking domains, similar to EFF’s Privacy Badger (light on info) and UW’s TrackingObserver (github, NSDI paper).
-   * Status: Mostly research oriented right now, none of these have been productionized or used at scale.
+** Status: Mostly research oriented right now, none of these have been productionized or used at scale.
-   * Roadblocks:
-      * (Research) None of these have been run on a large enough data sets to determine realistic false positive and false negative rates for users in the wild.
-      * (Policy) Per-user classification is not very interesting -- to take advantage of the Firefox user-base we’d want willing participants to share their data in order to improve global identification of tracking domains for everyone. Any feature that relies on user-input for correction or aggregation (such as the “Report Spam” feature in webmail) will require Mozilla to collect user data. Historically this has been challenging for us.
-      * (Implementation) Feasibility is not clear. Avoiding false positives (domains incorrectly identified as tracking) is much more important than avoiding false negatives (tracking domains not identified), but both are important for effectiveness and good UX. This requires running a service. Historically we don’t have a good track record on this kind of problem.