Intent to Extend Deprecation Trial: Restrict "private network requests" for subresources from public websites to secure contexts.

[Yifan Luo]

Contact emails

tit...@chromium.org, cl...@chromium.org, mk...@chromium.org, va...@chromium.org, l...@chromium.org

Explainer

https://github.com/WICG/private-network-access/blob/master/explainer.md

Specification

https://wicg.github.io/private-network-access

Design docs

https://docs.google.com/document/d/1x1a1fQLOrcWogK3tpFBgQZQ5ZjcONTvD0IqqXkgrg5I/edit#heading=h.7nki9mck5t64

Summary

Requires that private network requests for subresources from public websites may only be initiated from a secure context. Examples include internet to intranet requests and internet to loopback requests. This is a first step towards fully implementing Private Network Access: https://wicg.github.io/private-network-access/

Blink component

Blink>SecurityFeature>CORS>PrivateNetworkAccess

TAG review

https://github.com/w3ctag/design-reviews/issues/572

TAG review status

Issues addressed

Risks

Interoperability and Compatibility

No interoperability risks. Compatibility risk is small but non-negligible. UseCounters show ~0.1% of page visit making use of this feature. Direct outreach to the largest users per UKM data revealed no objections to this launch. Rolling this deprecation out to beta per the previous I2S resulted in more feedback about the compatibility risk and the need for a time extension. See the following doc for an extensive discussion: https://docs.google.com/document/d/1bpis0QwaA9ZrRFmpPW6LiaPmdwT0UhhUMNsEnU0zfLk/edit

Gecko: Worth prototyping (https://github.com/mozilla/standards-positions/issues/143) Tentatively positive, but no formal position yet.

WebKit: Positive (https://lists.webkit.org/pipermail/webkit-dev/2021-May/031837.html)

Web developers: Mixed signals (https://docs.google.com/document/d/1bpis0QwaA9ZrRFmpPW6LiaPmdwT0UhhUMNsEnU0zfLk/edit) In our recent survey, most of websites are able to migrate if our new permission prompt can be landed as a way for them to relax mixed content checks. https://docs.google.com/spreadsheets/d/1z5ZdCslNCnSVR7TNlUTHjSvunMFmT_9G9NOx8-O78-I/edit?resourcekey=0-DITlG8tDuFDWHiBUHnlSoQ#gid=309953809 ------------ Some websites, broadly falling in the category of controller webapps for IoT devices, find this change incompatible with their use cases. While many use cases can be solved with specific workarounds, some still require further engagement.

Other signals:

Activation

Developers of non-secure sites that rely upon local servers will need to upgrade to HTTPS. This might cause some complications, as mixed-content checks will begin to apply. Chrome carves out HTTP access to loopback (as perhttps://w3c.github.io/webappsec-secure-contexts/#localhost), which is a release valve for folks who don't want to go through the effort of securely-distributing certs for local servers. The initial launch in M92 was delayed due to compatibility risks surfaced during the rollout to beta. See this doc for a lot more details: https://docs.google.com/document/d/1bpis0QwaA9ZrRFmpPW6LiaPmdwT0UhhUMNsEnU0zfLk/edit

Security

This change should be security-positive.

WebView application risks

Does this intent deprecate or change behavior of existing APIs, such that it has potentially high risk for Android WebView-based applications?

Goals for experimentation

User feedbacks collection: https://docs.google.com/spreadsheets/d/1z5ZdCslNCnSVR7TNlUTHjSvunMFmT_9G9NOx8-O78-I/edit?usp=sharing&resourcekey=0-DITlG8tDuFDWHiBUHnlSoQ ------------ It seems that many developers have not noticed the upcoming launch despite outreach efforts, and will likely only notice once Chrome ships the secure context restriction. Thus delaying the launch by a few milestones to offer more breathing room to the currently-aware developers would not mitigate the risk when we ship the next time. A Deprecation Trial seems like the logical next step. This would allow us to protect the vast majority of users of the web by at least requiring attackers to sign up for the trial, itself a deterrent. Simultaneously, it would give enough time to legitimate websites to work around the new restriction. Finally, it would allow more time for discussions should our planned solutions fail to adequately address developers’ concerns.

Reason this experiment is being extended

We have collected 20+ developers' feedback since the last milestone. 85.7% developers said that they are still migrating to HTTPS, 50% said they need more time and 50% said they are not able to migrate local devices for various reasons and need future help. In the meanwhile, we are also collecting developers' feedback on our future plan for websites that cannot migrate their private devices to HTTPS but would like to migrate their public websites. 11.1% websites answered probably yes to our new feature and 72.2% responded might or might not. The major considers are they also need the allowance on frames/iframes (Q8 64.7%), want to use IP address as ids in permission (Q12 82.3%), too many permission prompt might be a spam (2 answers) and need to wait for other browsers supporting Private Network Access. In this case, we are also actively changing our further plan and collaborating with other browsers at the same time. ------------ The main workaround suggested to impacted websites was to use WebTransport's serverCertificateHashes feature. That is only shipping in Chrome 100; developers need more time to try it out. In addition, some issues have been identified with WebTransport that are prompting us to re-evaluate alternatives. In the meantime, keeping the trial going helps "staunch the bleeding" and provides a channel for discussing plans with affected web developers.

Ongoing technical constraints

None.

Debuggability

When a request is made that violates this restriction and the feature is not enabled, three things happen: 1. A warning message is logged to the DevTools console. 2. A deprecation report is filed against the initiator website's Reporting API, if so configured. 3. An issue is surfaced in the DevTools Issues panel. Likewise, when the feature is enabled and a request is blocked, the same happens except that the message logged to the DevTools console is an error and its text is slightly different. The devtools network panel shows information about the source and remote address spaces at play.

Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, Chrome OS, Android, and Android WebView)?

Yes

Is this feature fully tested by web-platform-tests?

Yes

Flag name

BlockInsecurePrivateNetworkRequests

Requires code in //chrome?

False

Tracking bug

https://crbug.com/986744

Launch bug

https://crbug.com/1129801

Estimated milestones

OriginTrial desktop last	113
OriginTrial desktop first	94
DevTrial on desktop	86

OriginTrial Android last	113
OriginTrial Android first	94
DevTrial on Android	86

Link to entry on the Chrome Platform Status

https://chromestatus.com/feature/5436853517811712

Links to previous Intent discussions

Ready for Trial: https://groups.google.com/a/chromium.org/g/blink-dev/c/EeGg7TxW6U4/m/7ZvqAqHLAwAJ
Intent to Experiment: https://groups.google.com/a/chromium.org/g/blink-dev/c/vlDZXlPb00k/m/1421ACiuAAAJ
Intent to Extend Experiment: https://groups.google.com/a/chromium.org/g/blink-dev/c/JPD001kqeck
Intent to Ship: https://groups.google.com/a/chromium.org/g/blink-dev/c/cPiRNjFoCag/m/DxEEN9-6BQAJ

This intent message was generated by Chrome Platform Status.

--

Yifan

[Yoav Weiss]

Can you clarify the deprecation trial timeline so far? The "intent to extend experiment" you linked to indicates experimentation until M106. Is that correct? 

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAG-zKU_1bPpOg-fexyo%2BXtMP%3D5_M4eKyeTPyWMD07EvUarogUA%40mail.gmail.com.

[Jonathan Hao]

Hi Yoav,

Here's the "intent to extend experiment" thread to M109. https://groups.google.com/a/chromium.org/g/blink-dev/c/PCfDnIV1cOw/

Best,

Jonathan

[Yoav Weiss]

Thanks for the link! :)

Do y'all have usage stats trends and/or insights into how far we are from being able to turn off the deprecation trial? 

[Yifan Luo]

I'm currently requesting opt-in/opt-out data from origin trial team and see how's the data looks like. Will send it in the thread after I have it.

The situation is that some of the local devices are not able to migrate to HTTPS. In this case, we introduced a permission prompt in the middle to relax mixed-content checks here and our current expectation is to ship it on M111.

However, some websites recently reached us and said even we shipped it, if the other browsers don't, it is still hard, even not possible, for them to migrate their public websites to HTTPS.

I'm currently writing the permission part of the spec and asking for positions from other browser vendors. ( I talked with anne@ from webkit once about it and he was at least interested in the spec at the time ) Even though, it might take long for all the browsers be able to ship it and as though websites able to migrate it.

So that I'm not able to give you any estimation on it at this point, but hopefully we can after gain the position from other browser vendors on it.

best,

Yifan

[Yoav Weiss]

LGTM to extend the deprecation trial till M113.

On Tue, Dec 13, 2022 at 3:11 PM Yifan Luo <l...@google.com> wrote:

I'm currently requesting opt-in/opt-out data from origin trial team and see how's the data looks like. Will send it in the thread after I have it.

The situation is that some of the local devices are not able to migrate to HTTPS. In this case, we introduced a permission prompt in the middle to relax mixed-content checks here and our current expectation is to ship it on M111.

However, some websites recently reached us and said even we shipped it, if the other browsers don't, it is still hard, even not possible, for them to migrate their public websites to HTTPS.

That's indeed a tricky situation. I wonder if we could add some form of content negotiation (e.g. a request header) that would enable such sites to redirect to HTTPS in the interim. Or maybe even tell them to use UA-CH for that purpose..

 

I'm currently writing the permission part of the spec and asking for positions from other browser vendors. ( I talked with anne@ from webkit once about it and he was at least interested in the spec at the time ) Even though, it might take long for all the browsers be able to ship it and as though websites able to migrate it.

Getting positions from other browsers sounds like a good next step.

[Guokai Han (Jackie)]

Late to the party. My thoughts are as follows:

1. This proposal introduces a lot of complexity to other parts, like proxy application, service worker, CORS.
2. It is difficult to upgrade all LAN services to https or add new CORS headers, or it will take many years.
3. This limits the capabilities of web applications. For example, a public PWA Web console app, which allows users to configure the local service address(and username/password) in a LAN network by users themselves, and then communicate with it. For example, various services deployed in a NAS device.
4. The original intention of this proposal is to protect LAN services, but I think the right direction should be to make these LAN services improve their own security.

Jackie