-
Notifications
You must be signed in to change notification settings - Fork 387
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fill out section on trusted UI #875
Changes from 2 commits
772daca
3bcdfe8
effd538
992c094
ffaa000
2e45ad9
b36b143
4a0b358
91a9c44
dd9065c
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -2214,25 +2214,47 @@ Note: Is is suggested that poses reported relative to a {{XRReferenceSpaceType/" | |
|
||
Note: Is is suggested that poses reported relative to a {{XRBoundedReferenceSpace}} be [=limiting|limited=] to a distance of 1 meter outside the {{XRBoundedReferenceSpace}}'s [=native bounds geometry=]. | ||
|
||
<section class="unstable"> | ||
Gaze Tracking {#gazetracking-security} | ||
------------- | ||
|
||
While the API does not yet expose eye tracking capabilities a lot can be inferred about where the user is looking by tracking the orientation of their head. This is especially true of XR devices that have limited input capabilities, such as Google Cardboard, which frequently require users to control a "gaze cursor" with their head orientation. This means that it may be possible for a malicious page to infer what a user is typing on a virtual keyboard or how they are interacting with a virtual UI based solely on monitoring their head movements. For example: if not prevented from doing so a page could estimate what URL a user is entering into the user agent's URL bar. | ||
|
||
To prevent this risk the user agent MUST set the [=visibility state=] of all {{XRSession}}s to {{XRVisibilityState/"hidden"}} when the user is interacting with sensitive, trusted UI such as URL bars or system dialogs. Additionally, to prevent a malicious page from being able to monitor input on other pages the user agent MUST set the {{XRSession}}'s [=visibility state=] to {{XRVisibilityState/"hidden"}} if the [=currently focused area=] does belong to the document which created the {{XRSession}}. | ||
|
||
Trusted Environment {#trustedenvironment-security} | ||
------------------- | ||
|
||
If the virtual environment does not consistently track the user's head motion with low latency and at a high frame rate the user may become disoriented or physically ill. Since it is impossible to force pages to produce consistently performant and correct content the user agent MUST provide a tracked, trusted environment and an [=XR Compositor=] which runs asynchronously from page content. The compositor is responsible for compositing the trusted and untrusted content. If content is not performant, does not submit frames, or terminates unexpectedly the user agent should be able to continue presenting a responsive, trusted UI. | ||
The user agent MUST support showing a <dfn>Trusted UI</dfn>, that is, an interface that the user can trust comes from the user agent, which the user may interact with without interference from the page. Some form of [=trusted UI=] MUST be used to show permissions prompts. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Nit: there's something about this sentence that seems somewhat oddly phrased, but I can't put my finger on how it might be clearer.
NellWaliczek marked this conversation as resolved.
Show resolved
Hide resolved
|
||
|
||
Additionally, page content has the ability to make users uncomfortable in ways not related to performance. Badly applied tracking, strobing colors, and content intended to offend, frighten, or intimidate are examples of content which may cause the user to want to quickly exit the XR experience. Removing the XR device in these cases may not always be a fast or practical option. To accommodate this the user agent SHOULD provide users with an action, such as pressing a reserved hardware button or performing a gesture, that escapes out of WebXR content and displays the user agent's trusted UI. | ||
|
||
When navigating between pages in XR the user agent should display trusted UI elements informing the user of the security information of the site they are navigating to which is normally presented by the 2D UI, such as the URL and encryption status. | ||
A [=trusted UI=] which does not exit immersive mode is known as a <dfn>trusted immersive UI</dfn>. User agents are not required to support [=trusted immersive UI=], they may instead temporarily pause/exit immersive mode to handle prompts to the user. | ||
NellWaliczek marked this conversation as resolved.
Show resolved
Hide resolved
|
||
|
||
A [=trusted UI=] MUST have the following properties: | ||
|
||
|
||
- It must not be spoofable | ||
- It indicates where the request/content displayed originates from | ||
- It doesn't rely on a shared secret that can be observed by a mixed reality capture (e.g. a gesture that can be seen by the camera) | ||
NellWaliczek marked this conversation as resolved.
Show resolved
Hide resolved
|
||
- It is consistent between immersive experiences in the same UA | ||
|
||
<div class="note"> | ||
Note: Examples of [=trusted UI=] include: | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Given that the text immediately above is specifically talking about trusted immersive ui, it took me a moment to realize this was back to being examples of the more general definition. |
||
- The default 2D mode browser shown when not in immersive mode | ||
- A prompt shown within immersive mode which can only be interacted with via a reserved hardware button to prevent spoofing | ||
- Pausing the immersive session and showing some form of desktop environment in which a prompt can be shown | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more.
I feel like the phrasing on this one is maybe a tad bit too restrictive? Specifically the use of the phrase "desktop environment", which seems to suggest a very specific and potentially mobile-excluding concept. What about something along the lines of this?
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. done |
||
|
||
</div> | ||
|
||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Did we also want to enumerate the properties of a trusted UI from #718 There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. done! |
||
|
||
|
||
If the virtual environment does not consistently track the user's head motion with low latency and at a high frame rate the user may become disoriented or physically ill. Since it is impossible to force pages to produce consistently performant and correct content the user agent MUST provide a tracked, trusted environment and an [=XR Compositor=] which runs asynchronously from page content. The compositor is responsible for compositing the trusted and untrusted content. If content is not performant, does not submit frames, or terminates unexpectedly the user agent should be able to continue presenting a responsive, [=trusted UI=]. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. This seems unrelated enough to trusted ui that it probably warrants its own section heading. |
||
|
||
Additionally, page content has the ability to make users uncomfortable in ways not related to performance. Badly applied tracking, strobing colors, and content intended to offend, frighten, or intimidate are examples of content which may cause the user to want to quickly exit the XR experience. Removing the XR device in these cases may not always be a fast or practical option. To accommodate this the user agent SHOULD provide users with an action, such as pressing a reserved hardware button or performing a gesture, that escapes out of WebXR content and displays the user agent's [=trusted UI=]. | ||
Manishearth marked this conversation as resolved.
Show resolved
Hide resolved
|
||
|
||
{{XRSession}}s MUST have their [=visibility state=] set to {{XRVisibilityState/"hidden"}} when the user is interacting with potentially sensitive UI from the user agent (such as entering a URL) in the trusted environment. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Is this a duplicate of the last paragraph in the gaze tracking section? |
||
|
||
<section class="unstable"> | ||
Gaze Tracking {#gazetracking-security} | ||
NellWaliczek marked this conversation as resolved.
Show resolved
Hide resolved
|
||
------------- | ||
|
||
While the API does not yet expose eye tracking capabilities a lot can be inferred about where the user is looking by tracking the orientation of their head. This is especially true of XR devices that have limited input capabilities, such as Google Cardboard, which frequently require users to control a "gaze cursor" with their head orientation. This means that it may be possible for a malicious page to infer what a user is typing on a virtual keyboard or how they are interacting with a virtual UI based solely on monitoring their head movements. For example: if not prevented from doing so a page could estimate what URL a user is entering into the user agent's URL bar. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Is this text still helpful?
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. This text can probably be repurposed as the intro paragraph if it is adjusted slightly.
|
||
|
||
To prevent this risk the user agent MUST set the [=visibility state=] of all {{XRSession}}s to {{XRVisibilityState/"hidden"}} when the user is interacting with sensitive, trusted UI such as URL bars or system dialogs. Additionally, to prevent a malicious page from being able to monitor input on other pages the user agent MUST set the {{XRSession}}'s [=visibility state=] to {{XRVisibilityState/"hidden"}} if the [=currently focused area=] does belong to the document which created the {{XRSession}}. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. We should make it clear that this applies to both types of trusted UI |
||
|
||
Context Isolation {#contextisolation-security} | ||
----------------- | ||
|
||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It might be a good idea to put a couple sentences before this elaborating on why trusted UI is a harder problem in immersive experiences (i.e. all pixels can be drawn by the developer which allows for spoofing).