Privacy/Features/Shortened HTTP Referer header

From MozillaWiki
Jump to: navigation, search
Please use "Edit with form" above to edit this page.


Shortened HTTP Referer header
Stage Development
Status `
Release target `
Health OK
Status note `


Product manager Sid Stamm
Directly Responsible Individual Sid Stamm
Lead engineer Sid Stamm
Security lead Curtis Koenig
Privacy lead Sid Stamm
Localization lead `
Accessibility lead `
QA lead Mihai Morar
UX lead `
Product marketing lead `
Operations lead `
Additional members Owen Chu

Open issues/risks

  • Site breakage (tolerable? need to quantify this)
  • User confusion (make it hidden pref?)

Stage 1: Definition

1. Feature overview

There is the desire to remove the Referer header outright, possibly in favor of the Origin header or something with less information. It can leak sensitive data accidentally and can be abused as a form of ambient authority. Unfortunately, we can't just stop sending it on requests because too many things on the web might break. Sites like Facebook have expressed a desire for a referer-shortening capability, so this feature benefits both users and site developers.

This feature adds a way to attenuate the information that's sent as the referrer. This is multiple phases:

Phase 1: Plumbing for global and site-specific control. In the first phase, we will create a global pref so power users and addons can select at most how much of the URL is sent as referrer, globally. Additionally, on a site-by-site basis (probably via permission manager) a referrer policy can be applied to override the global setting.

Phase 2: Site-based control via CSP header and tag attributes. In the second phase, we enable sites to reduce the amount of data transmitted in referrers generated on their site. This is done by the site sending a signal with Content Security Policy response header indicating that outgoing referrers should be reduced. Stripping options should include the same options mentioned in phase 1. Sites will be able to specify a default policy for their pages, but we will also implement some per-link control (One mechanism could be to support the rel="noreferrer" attribute to omit referers from link clicks, or similar).

Phase 3: Site-based control via meta tag. Once the CSP HTTP response header can be used to set a site policy, we will extend CSP to be settable via the meta tag. This way sites can specify a referrer policy without having to send an HTTP header.

Note: Site-based control could be accomplished using the meta referrer tag, but we are opting to implement the same functionality in CSP to combine the security/privacy features. See whatwg wiki and see also bug 704320.

Phase 4: New Firefox defaults. Once sites have better control, we will decide how much to limit referrer sending by default (same-origin and cross-origin). Right now too much information is transmitted in referer headers, and we should reduce that to the extent possible. In this phase, we'll have an open discussion in dev-privacy about what defaults to choose.

Phase 5: Extra bonus phase - UI for users to control how much referrer is sent on a global basis, likely in the privacy settings for Firefox. Additionally, per-site control in the permissions and siteinfo UI. This may not be necessary if Phase 4 is successful and there's not much referrer sending by default.

See also: bug 587523

2. Users & use cases

Nice document from Dan Aurbach:

Leaking search terms 
From bug 587523#c0: "An example of this can be seen by searching for 'no knead bread' with Google, and clicking on the 4th search result, which takes you to, a page which "helpfully" lets you know that it is aware of the search terms that brought you to the site."
Outbound link anonymization 
Many sites like gmail send outbound links through a common redirect to strip off any information that may be present in the URL. Supporting rel="noreferrer" reduces the need for extra HTTP traffic and redirects.

3. Dependencies


4. Requirements

  • Test plan must be created and implemented
  • Use cases must be clearly outlined and it must be clear how the feature addresses each.
  • Initially, Phase 1 (user-set) should not change default behavior until user initiates change.
  • Default referer behavior for sites should not change until sites activate attenuation features.


  • We are not removing the HTTP referer header
  • We are not replacing the HTTP referer header
  • This is not the Origin header

Stage 2: Design

5. Functional specification

See also the noreferrer link type

6. User experience design


Stage 3: Planning

7. Implementation plan

Phase 1:

  • [DONE] global default referrer policy via pref (bug 822869)
  • [ON TRACK] site-specific setting via permission (bug 966505)

Phase 2:

  • [DONE] CSP referrer directive support (bug 965727)
  • [DONE] noreferrer attribute for anchor tags (bug 530396)
  • [NEW] referrer (policy) attribute for anchor tags

Phase 3:

Phase 4:

  • [NEW] discussion in dev-privacy about measuring success of new referrer defaults
  • [NEW] discussion in dev-privacy to choose new policy (and measurement of its effects)
  • [NEW] change global default policy for referrer bug 122453

Phase 5:

  • [NEW] decide if this phase should be dropped or completed
  • [NEW] about:permissions and siteInfo UI for per-site referrer settings
  • [NEW] privacy settings pane UI for global referrer settings

8. Reviews

Security review


Privacy review


Localization review




Quality Assurance review


Operations review


Stage 4: Development

9. Implementation

See above in Planning section for progress. Just some links here.

Stage 5: Release

10. Landing criteria

  • Has tests
  • Tests pass on B2G, Android and Desktop

Feature details

Priority P1
Rank 4
Theme / Goal Tracking Control
Roadmap Privacy
Secondary roadmap `
Feature list Platform
Project `
Engineering team Networking

Team status notes

  status notes
Products ` `
Engineering ` `
Security ` `
Privacy ` `
Localization ` `
Accessibility ` `
Quality assurance ` `
User experience ` `
Product marketing ` `
Operations ` `