Progress in the Privacy Sandbox (May - June 2022)

Welcome to this edition of Progress in the Privacy Sandbox, covering May and June 2022, as we track the milestones on the path to phasing out third-party cookies in Chrome and working towards a more private web. In each edition, we share an overview of the updates and news across the Privacy Sandbox.

Privacy Sandbox Relevance and Measurement origin trial

We're continuing to run the combined origin trial for Attribution Reporting, FLEDGE, and Topics, along with expanding it to include Fenced frames and Shared Storage. Fenced frames provide a restricted container for displaying content which FLEDGE makes use of to display ads. Shared Storage complements fenced frames by allowing access to a carefully gated form of unpartitioned storage where the stored data may be used as part of the ad selection process.

The origin trial has now been expanded to 50% of Chrome Beta users. At this stage, the focus is still on testing the infrastructure setup, developer experience, and user interface, before moving on to larger scale effectiveness or utility testing. In this testing we discovered and fixed a crashing bug in Topics where a null value was not handled correctly. In these early stages of testing, it's common to find issues as we start running real traffic through the feature. This is one of the reasons we start with a small portion of overall traffic to still discover issues while minimizing the impact of them. However, as this had the potential to crash the browser for that small set of users, we disabled the Topics API in the origin trial while the fix rolled out. This fix is now fully complete, and Topics is enabled again within the origin trial.

Your feedback and testing at this stage is vital to ensure we're proposing and building the functionality you need. If you participate in the origin trial now, you can expect to see continued regular updates to the code as we respond to feedback, issues, and expand the available functionality.

You can sign up for the origin trial now. We have full instructions on how to join, how to test, demos to explore, and where to provide feedback for the different aspects of the trial.

Feedback

Feedback from a diverse set of stakeholders across the web ecosystem is critical to the Privacy Sandbox initiative as a whole. The dedicated feedback section provides an overview of the existing public channels, where you can follow or contribute to discussion, along with a feedback form to ensure you can always reach the Chrome team directly.

We compiled a summary of your feedback and our responses in a 2022 Q1 feedback overview report. There's a lot there—so you'd be forgiven for not reading the whole thing! Hopefully though you can get a feeling for the types of questions and issues being raised along with how we then deal with those. We aim to make these feedback routes easy and accessible for developers, so if you have a question or something to clarify, then we want to hear it.

We're continuing to run our series of Privacy Sandbox Office Hours with an overview session on Attribution Reporting. This is your chance to ask questions directly to the implementation team. We're also planning to run a version of this session in Japanese, and another session for an updated demo walk-through. There will be further editions covering other APIs and aspects of the project which we will publicize across the mailing lists, blogs, and Twitter.

Strengthen cross-site privacy boundaries

Third-party cookies are a key mechanism that enables cross-site tracking. Being able to phase them out is a major milestone, but we also need to address other forms of cross-site storage or communication.

Cookies

As the cookie-related proposals progress, you should audit your own SameSite=None or cross-site cookies and plan the action you will need to take on your site.

CHIPS

CHIPS (Cookies Having Independent Partitioned State) allows developers to opt a cookie into "partitioned" storage, with a separate cookie jar per top-level site. We are extending the current origin trial through to the end of Chrome 104 around the end of August. You can sign up for the CHIPS origin trial now and we have developer instructions available so you can test cookies with the Partitioned attribute on your own production site.

We are also continuing to clean up and improve the general default functionality of cookies alongside the changes that are under the Privacy Sandbox banner. We've sent the I2P (Intent to Prototype) for Origin-bound cookies (by default), which would result in more secure defaults for a cookie. While the earlier SameSite=Lax by default change meant cookies were restricted to the same site (or "first-party") by default, they can still be sent across different ports or URL schemes. This update would mean cookies are only sent to the exact origin on which they were set unless explicitly allowed by the Domain attribute.

We're also bringing Chrome more inline with the Fetch specification by blocking the Set-Cookie header on outbound fetch requests. Reporting shows that usage of this ability is extremely low, but you should be aware that if you rely on this functionality, those cookies will soon stop being set.

Shared Storage

Shared Storage allows for sites to store unpartitioned data but only read that data back in a secure environment with carefully constructed output gates. Shared Storage pairs with fenced frames which provides the secure environment allowing use cases such as A/B testing over a campaign.

The I2E for Shared Storage makes it available for testing as part of the wider Privacy Sandbox Relevance and Measurement origin trial. Developer documentation is available covering use cases and testing.

Preventing covert tracking

As we reduce the options for explicit cross-site tracking, we need to address the areas of the web platform that expose identifying information that enables fingerprinting or covert tracking of users.

User-Agent string reduction and User-Agent Client Hints

We continue to incrementally reduce the information passively available in Chrome's user-agent string and providing alternative User-Agent Client Hints (UA-CH) for sites that need to actively request that information. The initial phase of reducing the minor version number to zeroes is now fully rolled out to Chrome 101 and above.

Old

Mozilla/5.0 (Linux; Android 12; Pixel 6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/101.0.4638.16 Mobile Safari/537.36

New

Mozilla/5.0 (Linux; Android 12; Pixel 6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/101.0.0.0 Mobile Safari/537.36

We are continuing to refine and update the general Client Hints behavior. This includes cleaning up the default for legacy hints (dpr, width, viewport-width, and device-memory) so they are not sent to third-party subresources by default.

Accept-Language reduction

The Accept-Language header sends a user's language preferences to a site, which is useful for delivering localized content but can represent a source of passive fingerprinting information—especially when the user has multiple accepted languages. We sent an I2P to reduce the fingerprinting surface in the Accept-Language header.

The intent is that on the first request to a site the browser would only send the top preferred language in the header, such as Accept-Language: fr. The site response should then specify the Content-Language of the response and indicate if there are multiple languages available by using the Vary and Variants headers. For example:

Get / HTTP/1.1
Host: example.com
Accept-Language: fr

HTTP/1.1 200 OK
Content-Language: fr
Vary: Accept-Language
Variants: Accept-Language=(de en fr)

The browser can use the Variants information to re-request content in a preferred language in the event the top choice is not available. I2Ps are deliberately sent early in the process to allow discussion of the potential impact of a change like this and ensure that we set up appropriate metrics to monitor that impact.

Fenced frames

A fenced frame (<fencedframe>) is a proposed HTML element for embedded content, similar to an iframe. Unlike iframes, a fenced frame restricts communication with its embedding context to allow the frame access to cross-site data without sharing it with the embedding context. For example, in FLEDGE the intent is for ads to be displayed within a fenced frame.

You can read the new developer overview content. We published the I2E for fenced frames and you can now sign up as part of the wider Privacy Sandbox Relevance and Measurement origin trial.

Show relevant content and ads

As we move towards phasing out third-party cookies, we are introducing APIs that enable key use cases that sites depended on to allow them to fund their content without continuing to enable cross-site tracking.

Topics

The Topics API is a proposal to enable interest-based advertising without cross-site tracking. We sent an I2E to include Topics as part of the Privacy Sandbox Relevance and Measurement origin trial. We also have new developer guidance for the testing and providing feedback on Topics during the origin trial.

As this is early stage testing, we are actively discovering and addressing issues in the code as they come up. On Topics, we discovered a crashing bug, so we temporarily disabled the API within the origin trial roll out the fix without overly impacting the user experience. With that fix complete, Topics API is now enabled for 50% of Chrome Beta users as part of the overall origin trial.

FLEDGE

FLEDGE enables remarketing and custom audience use cases, as in advertising that can make use of sites or products previously visited, without relying on an individual identifier. We sent an I2E for FLEDGE, again to enable it as part of the wider Privacy Sandbox Relevance and Measurement origin trial. And likewise, there's matching developer documentation for the experiment available.

Measure digital ads

As a companion to display ads without cross-site tracking, we need privacy-preserving tools to measure the effectiveness of those ads.

Attribution Reporting API

The Attribution Reporting API allows adtech and advertisers to measure events on one site, like clicking or viewing an ad, that lead to a conversion on another site—without enabling cross-site tracking. As you may have guessed, there was also an I2E for Attribution Reporting to continue expanding its testing as part of the Privacy Sandbox Relevance and Measurement origin trial.

During the initial stage of the origin trial we are focused on feedback around the developer experience and integration, such as debugging, and this will expand to cover end-to-end testing across event-level and summary reports.

Article feedback

As we continue to publish these updates and progress through the Privacy Sandbox as a whole, we want to make sure that you as a developer are getting the information and support that you need. Let us know on @ChromiumDev Twitter if there's anything that we could improve in this series. We'll use your input to continue improving the format.