Private State Token API

[hober]

Request for Mozilla Position on an Emerging Web Specification

• Specification Title: Trust Token API
• Specification or proposal URL: https://wicg.github.io/trust-token-api/
• Explainer: https://github.com/WICG/trust-token-api
• Mozillians who can provide input (optional):

Other information

JS API built on top of Privacy Pass, see #261.

[martinthomson]

See #261 for our assessment of the underlying protocol. We are deferring assessment there and so will defer here also.

[martinthomson]

Google deployed the Privacy Pass protocols on the Web under the name “Private State Tokens”, which has been discussed in the Anti-Fraud Community Group of the W3C. We see several serious flaws in this deployment that mean that we have to take a “negative” position.

There are several characteristics of Google’s implementation of Privacy Pass for the Web that deserve praise. In particular, the detailed specification for Web integration is appreciated. We also appreciate the effort to support the provision of tokens by arbitrary sites, which goes some way to address our concerns about equitable access for sites.

However, Google’s proposal and deployment do not place adequate controls on the semantics of tokens. Having a more open policy for participation by sites means that there is no obvious way to ensure that tokens carry acceptable semantics.

Google requires that sites make a public declaration (https://github.com/privacysandbox/attestation) that they will not use this API for tracking purposes, but we do not believe that this is an adequate measure. There are potential abusive uses of tokens that do not involve tracking.

The choice to use a joint attester/issuer deployment model means that the issuer and service can use timing to correlate issuance (where private information is revealed) and redemption (where that information should not be).

The controls that we list in the latter half of our response to Apple’s proposal apply equally to Google’s proposal. In summary, we are skeptical that it is possible to develop adequate controls on the actions of issuers for this system to work. Our analysis goes into further detail.

We understand that the intent is to iterate on the proposal with a goal of eventually addressing these concerns, but we have to judge based on what is documented, implemented, and deployed.

[offizium-berndstorath]

@martinthomson

Google’s proposal and deployment do not place adequate controls on the semantics of tokens

Please elaborate

There are potential abusive uses of tokens that do not involve tracking.

Please elaborate

Our analysis goes into further detail.

Not really