Storage access policy: Block cookies from trackers

Pre-release versions of Firefox include a new storage access policy that blocks cookies and other site data from third-party tracking resources. This new policy is designed as an alternative to thecurrent policies offered in release versions of Firefox, and protects against cross-site tracking while minimizing the site breakage associated with cookie blocking. This article explains how the new policy works and how you can test it.

Testing in Nightly

If you’d like to test how your website works under the new policy we encourage you to download Firefox Nightly and try it out. You can enable it through the “Content Blocking” menu in theControl Center in Firefox Nightly:

• Download, install, and open upFirefox Nightly.
• Open the Content Blocking menu in the Control Center.
• Click "Add Blocking…" next to "Third-party Cookies".
• On the page that follows check "Third-party Cookies" and set the radio button to "Trackers".

If you find a website broken as a result of this change, file a bug under the Tracking Protection component within the Firefox product onBugzilla.

As this is an experimental policy, it will likely change as we discover site breakage and fix bugs. We’ll keep this page updated with the newest information.

Tracking protection explained

How does Firefox determine which resources are tracking resources?

Firefox uses the Tracking Protection list to determine which resources are tracking resources. The Tracking Protection list is maintained by Disconnect. When the list is applied in Firefox, we make two important changes:

• First, we only use the "Basic Protection" version of the list, which excludes some categories of trackers. In the future, we may expand our protections to use the "Strict Protection" version of the list.
• Second, Firefox will not classify domains as trackers when they are loaded on a top-level site owned by the same organization.

Firefox uses the built-inTracking Protection URL classifier to determine which resources match the tracking protection list. Domains are matched against the list in accordance with the SafeBrowsing v4 specification. Specifically, we check the exact hostname of the resource against the list, as well as the last four hostnames formed by starting with the last five components and successively removing the leading component. Consider the following examples:

Hostname on the list	Hostname of resource	Matched
example.com	example.com	Yes
example.com	a.b.example.com	Yes
blah.example.com	example.com	No
a.b.example.com	c.d.example.com	No
blah.example.com	foo.blah.example.com	Yes

What does the storage access policy block?

The storage access policy blocks resources identified as trackers from accessing their cookies and other site storage when they are loaded in a third-party context. This prevents those resources from retrieving tracking identifiers stored in cookies or site storage and using them to identify users across visits to multiple first parties. Specifically, Firefox does this by imposing the following restrictions:

Cookies:

• Block Cookie request headers and ignore Set-Cookie response headers.
• Return an empty string for calls to Document.cookie and ignore requests to set cookies via Document.cookie.

DOM Storage:

• localStorage: read and write attempts throw a SecurityErrorexception.
• sessionStorage: read and write attempts are permitted.
• IndexedDB: read and write attempts throw a SecurityErrorexception.

Messaging and Workers:

• Broadcast Channel: attempts to create a new BroadcastChannel will throw a SecurityErrorexception.
• Shared Worker: attempts to create a new SharedWorker will throw a SecurityErrorexception.
• Service Worker: attempts to create a new ServiceWorker will throw a SecurityErrorexception.

DOM Cache:

• Calls to CacheStorage will always reject with a SecurityError.

Browser caches:

• TheHTTP cache and the Image cache are partitioned for tracking resources, such that each top-level origin will have a separate partition and tracking resources on different top-level origins will be cached separate from each other.

What is not blocked by the policy?

1. This policy does not currently restrict third-party storage access for resources that are not classified as tracking resources. We may choose to apply additional restrictions to third-party storage access in the future.
2. The restrictions applied by the policy will not prevent third-party scripts classified as tracking resources from accessing storage in the main context of the page. These scripts can continue to use storage scoped to the top-level origin.
3. Origins classified as trackers will have access to their own storage when they are loaded in a first-party context.
4. Origins normally classified as trackers will not be blocked if the top-level page origin is determined to be from the same organization as them.

Storage access granted to tracking resources on first party basis

In order to improve web compatibility and permit third-party integrations that require storage access, Firefox will grant storage access scoped to the first party for a particular third-party origin under the following conditions. Currently, Firefox includes some web compatibility heuristics that grant storage access to third-party resources classified as trackers when a user interacts with those third parties. We do this when we expect that not granting access would result in the web page to break.  We’re also working on improving our support for the Storage Access APIas a way for embedded <iframe> s to request access. Going forward, we recommend third-parties to switch to using the Storage Access API in order to ensure being able to guarantee their access to storage when needed.

Automatic storage access upon interaction

In order to improve web compatibility, Firefox currently includes some heuristics to grant storage access automatically to third parties that receive user interaction. These heuristics are intended to allow some third-party integrations that are common on the web to continue to function. They are intended to be temporary and will be removed in a future version of Firefox. They should not be relied upon for current and future web development.

Third-party storage access may be granted to resources that have been classified as tracking resources when a user gesture triggers a pop-up window that hasopener access to the originating document. When that occurs, there are two possible ways a third-party origin can be granted access:

• The origin of the resource that is initially loaded in the pop-up window is granted storage access on the opener document.
• After the initial resource is loaded in the pop-up window, the window may go through a series of redirects to other hosts. If a user interacts with the pop-up window following a redirect, the origin of the content loaded in the pop-up window is given storage access on the opener document.

Scope of storage access

When storage access is granted, it is scoped to the origin of the opener document or subdomains of that origin. Access that is granted on the subdomain of an origin does not extend to the top-level origin. As an example, if a resource from tracker.exampleis granted storage access on foo.example.com, then tracker.examplewill be able to access its cookies on bar.foo.example.combut not example.com. Instead, if tracker.examplewere granted access on example.comit would be able to access its storage on bar.foo.example.com, foo.example.com, and example.com.

When storage access is granted to tracker.exampleon example.com, all resources loaded from tracker.exampleon any top-level document loaded from example.comare immediately given storage access. This includes all resources loaded in the main context of the page, embedded <iframe>s, and resources loaded within embedded <iframe>s. Storage access is not extended to other resources loaded on example.com(e.g. other-tracker.example), nor to other first parties on which tracker.exampleis embedded (e.g. example.org).

Storage access grants extend into the first level of nested contexts, but no further. This means that <iframe>s embedded in the main context of the page and loaded from a domain classified as a tracker will have full access to all storage locations accessible through JavaScript. Similarly, requests for resources loaded in <iframe>s embedded in the main context of the page will have access to HTTP cookies. However, further nested contexts, including but not limited to those from the origin classified as a tracker, will not be granted storage access.

Consider the following embedding scenarios on a top-level page loaded from example.comon which tracker.examplehas been granted storage access.

Embedding	tracker.example resource storage access
An image is loaded from tracker.exampleand embedded in the main context of example.com.	HTTP: Yes JS: N/A
example.comembeds an <iframe>from example.org. That <iframe>goes on to load an image from tracker.example.	HTTP: Yes JS: N/A
example.comembeds an <iframe>from example.org. That <iframe>goes on to embed an <iframe>from tracker.example.	HTTP: Yes JS: No
example.comembeds an <iframe>from tracker.example.	HTTP: Yes JS: Yes
example.comembeds an <iframe>from example.com(same origin). The nested <iframe>embeds an <iframe>from tracker.example.	HTTP: Yes JS: No

Storage access expiration

The storage access grant expires after 30 days. Domains classified as tracking resources may be granted third-party storage access on multiple first parties, and the storage permission for each party expires independently. The above heuristics will also serve to extend the lifetime of a third-party storage permission on origins that have already been granted access.  Each time the heuristic is activated, or a success call to the Storage Access API is made, the pre-existing storage access expiration will be extended by 30 days, counting from the time the previous access was granted.

Please note that in the future we expect to make changes to how long storage access will remain valid for.  As mentioned before, the way to know that you will be able to use storage as a third-party going forward will be using the Storage Access API.

Debugging

We encourage site owners to test their sites, particularly those that rely on third-party content integrations. We’ve added several new features to Firefox to make testing easier.

Devtools notifications

The devtoolsNetwork Monitor now includes an indicator for all resource requests that have been classified as tracking resources. This indicator is shown as a shield icon in the domain column. In the sample image below, trackertest.orgis classified as a tracking resource, while the request to example.com is not.

Adding custom domains to the Tracking Protection list

Curious how things will work if a third-party domain on your site were classified as a tracker? We’ve added a preference that allows you to add custom domains to the Tracking Protection URL classifier. To do so:

about:config

Warning: Be sure to remove these entries after you have finished testing.

FAQ

This cookie policy has the potential to lead to site breakage, but has been designed to allow common third-party integrations to continue to work while preventing cross-site tracking. In this section we describe the functionality you can expect in different integration scenarios.

Will this storage access policy block ads from displaying on my website?

No — this feature only restricts access to cookies and site data that can be used to track users across websites. Blocking tracking identifiers does not prevent the display of advertisements.

I use a third-party analytics service that is classified as a tracker. Will I still receive analytics data?

This depends on how the third-party analytics service is implemented. Third-party analytics providers will no longer be able to user their third-party storage to collect data. This means that providers using cookies which are scoped to their third-party domain, or local storage and other site data stored under their origin, will no longer have access to those identifiers across other websites.

If these services are embedded into the main context of the page, they can continue to use first-party cookies and site storage to track users across page visits on that specific first-party domain.

I use third-party services for social login, like, and share button integration. Will my users still be able to make use of these services?

This depends on how the social integration is implemented. We expect that many of the popular social integrations will continue to function as they do under Firefox’s current cookie policy with some minor differences in the user experience.

A social content provider that is classified as a tracker will not have access to their third-party cookies when the user first visits a new first party. Thus, the user may appear logged out to the service despite being logged in when they visit the provider’s website directly. Depending on the type of integration, the user may have to take some action to interact with the social content provider before the provider is given access to their cookies. For example:

• For social login, the user may have to click a login button on the first party.
• For social like or share buttons, the user will have to first interact with the button in a logged-out state. Once they do, many social content providers will prompt them to log in.

After these interactions, the provider will receive third-party storage access if they prompt the user in a way that is captured by the storage access activation heuristics described above. These providers should consider switching to explicitly request storage access through the Storage Access API as soon as possible. An initial implementation of this APIis currently available in Nightly.