This document is an editors draft and currently does not reflect consensus of the WG but rather is a starting point for further work. It is based on input documents and list discussion.
Privacy considerations are important to Device APIs, since misuse of information exposed by the APIs can have potentially harmful financial, physical safety, reputational and other impacts. Privacy needs a systemic solution that includes functional requirements on user agents, web sites and other components of the system, since any opportunity for misuse of private information is a risk. Addressing privacy may include functional requirements in technical standards, laws and regulations, and best practices.
While privacy is an important consideration for all APIs, privacy risks may vary according to the information exposed by an individual API. For example, inappropriate disclosure of contacts or location information could create serious personal safety issues in a broad range of cases, while disclosure of certain system information might create privacy risks in fewer contexts.
Privacy is a broad topic with various aspects that involve different parties in a system. In order to be clear which aspects are addressed and how, we take an architectural approach of breaking down the problem into smaller pieces. This should provide clarity and enable privacy-respecting solutions that can be adopted. We have identified the following pieces:
This is similar to writing APIs in such a way that they are security-friendly; you can't keep people from making mistakes, but you can make it easier and more natural to be privacy-respecting through the API design. An example is supporting the concept of Minimization by designing APIs that return the minimum data needed for a task, such as only obtaining address fields when an address is needed.
User agents are crucial to ensuring that user privacy is protected, but this capability must take implementation and adoption considerations into account. User agent design decisions can be separated to a large degree from API design decisions.
This can be hard to manage technically but might be possible through a simpler approach of defining and agreeing upon a small set of privacy preferences, similar to Creative Commons copyright licenses, that users can attach to their data. Defining a simple vocabulary for privacy could enable privacy rulesets that can be referenced by URI. Other people have had this idea as well.
WAI and the Web community at large have done a great job raising awareness about accessibility issues, and while implementations are not perfect, their effort has had a very measurable impact. There is therefore experience to be tapped in such an approach for the parts of the problem that depend on convincing people to do the right thing (which in some cases can be wide-ranging, including making script libraries support the solution directly, or having various organizations enforce it internally).
Privacy protections are frequently understood as a set of principles or elements (one such set is described in [[PRIVACY-ISSUES-GEO]]). The core elements of privacy that are relevant to Device APIs, user agents that support the APIs, and applications that use the APIs are as follows:
Notice: Informing users about the data collected through Device APIs
Consent: Obtaining user agreement to sharing data through Device APIs
Minimization: Limiting the amount and level of detail of data collected through Device APIs
Control: Allowing users to control access to their data once they have consented to having it collected through Device APIs
Access: Providing users with access to information about the data that has been collected about them through Device APIs
Retention: Limiting how long applications retain data that was collected through Device APIs
Secondary Use: Limiting applications from using data collected through Device APIs for purposes other than the purpose for which it was collected
Sharing: Limiting applications from sharing data collected through Device APIs with third parties
These elements will each need to be approached in different ways. Approaches include specific requirements on individual APIs, conveying user expectations together with the data itself, and/or documenting best practices for application and content developers. Certain approaches are better suited to safeguarding certain privacy elements than others.
This document provides specific requirements for individual APIs, addressing the elements that are most relevant to API definitions: notice, consent, minimization, control, and access. Requirements involving user expectations, which primarily address retention, secondary use, and sharing, will be documented separately. Best practices for developers will also be documented separately, covering notice, minimization, control, access, retention, secondary use, and sharing in the application developer context.
The following table summarizes the breakdown of how each element is covered:
Privacy Element | Requirements for API Definitions | Requirements related to User Expectations of Data Use | Best practices for developers |
---|---|---|---|
Notice | X | X | |
Consent | X | X | |
Minimization | X | X | |
Control | X | X | |
Access | X | X | |
Retention | X | X | |
Secondary Use | X | X | |
Sharing | X | X |
The privacy requirements for individual APIs are provided in the next section. The requirements described in this document are intended to be applicable to device APIs both in the context of widgets and web applications.
Issue: The breakdown described above foreshadows the idea of providing API hooks that allow users to attach their expectations/preferences/policies about privacy to the data they share through the APIs. Attaching policy rules to the data that get shared can provide a legal basis for enhancing the control users have over their data once they are shared; but doing so create the following challenges:
Many of the requirements listed here are recommendations (SHOULDs) rather than absolute requirements (MUSTs). In many cases this is because making a requirement absolute is appropriate for only a subset of the APIs, but not every API. As appropriate, individual APIs may place stronger normative requirements on user agents than the requirements in this document place on APIs.
Making sure that users understand the implications of using an application that relies on a Device API is fundamental to ensuring the protection of their data. The following requirements can help to make sure that users are properly notified:
APIs MUST make it possible for user agents to notify users that their data is being collected via the API. This notification MUST identify the application (for example, by displaying its document origin [[HTML5]]) and the precise data being collected.
APIs SHOULD provide support for visual indicator(s) that data is being collected via the APIs.
Issue: Should the APIs have a hook for applications to convey the intended usage of the data? If they do, should it be a required parameter? And how can this information be conveyed without misleading the user about the trustworthiness of that information?
The semantics of some APIs are defined such that user interaction is essential to make use of the API. An example is a camera API that enables the user to take a photograph, but is defined to require the user to press a shutter key to take the photograph. Another example is an API that produces a message template, but requires the user to press "send" to actually send the message.
Such user actions constitute implicit consent to collection of data via the API, since the user has a choice to perform these actions and doing so implies consent for the application to access the associated Device Capabilities. In such situations where it is obvious that performing the action involves sharing data with the application and the application’s intended use of the data is also obvious, additional dialogs that prompt users for consent may not be necessary.
Device APIs may also be defined such that consent must be explicit, not implicit. Examples are a camera API that takes a photograph without user involvement, or a messaging API that sends a message without the user pressing "send." In these cases dialogs may be required.
To ensure that data is not collected without users knowing or realizing, APIs should be designed with the presumption that the explicit consent model will be used, and should explain the specific circumstances under which implicit consent may be acceptable. This gives rise to the following requirements:
APIs MUST make it possible for user agents to obtain user consent before sharing any data via the APIs.
APIs MUST be defined in such a way that explicit consent is assumed, and they SHOULD articulate the circumstances under which implicit consent is acceptable.
An important caveat to the consent model supported by Device APIs relates to data about other people that the user may have on his or her device and be able to share via the APIs (other people’s contact or calendar information, for example). A user should not be able to consent through the Device APIs to use of other people’s information beyond the original interaction with the API. Thus, for example, a user should be able to consent to have an application contact another person mentioned in a calendar entry (perhaps to say “I am running late”), but the user should not also be able to consent to have the application make later use of the person’s contact information (perhaps to send marketing messages to that person). This caveat should be conveyed where appropriate in the APIs, best practices, and other Device APIs documents.
To reduce the risks of over-exposing users data, it is important to design APIs so that Web developers can request as little information as they need to accomplish their goals.
An example use case is a social networking case where the contacts API is used to contacts who are also members of a social network. Email addresses serve as the social network handles. In this case limiting results to the addresses and not other personal information is an example of minimization.
This gives rise to the following requirements:
APIs MUST make it easy to request as little information as required for the intended usage.
For instance, an API call should require specific parameters to be set to obtain more information, and should default to little or no information.
APIs SHOULD make it possible for user agents to convey the breadth of information that the requester is asking for.
For instance, if a developer only needs to access a specific field of a user address book, it should be possible to explicitly mark that field in the API call so that the user agent can inform the user that this single field of data will be shared.
APIs SHOULD make it possible for user agents to let the user select, filter, and transform information before it is shared with the requester.
The user agent can then act as a broker for trusted data, and will only transmit data to the requester that the user has explicitly allowed.
Given the sensitivity of the data made available through Device APIs, it is important for users to be able to control which applications have access to that data. The following requirements ensure that (1) users have control over their data even after they have shared it with an application, and (2) users have robust controls over which applications can obtain their data to begin with:
APIs MUST make it possible for user agents to support the revokation of user consent to sharing of data via the APIs.
APIs SHOULD support the ability for user agents to allow users to whitelist trusted applications and blacklist untrusted applications.
Notice and control cannot be fully implemented unless users can review how they have shared data in the past. The following requirements suggest how APIs can support users’ access to this information:
APIs SHOULD make it possible for user agents to allow users to view the applications with whom they have shared data.
APIs SHOULD make it possible for user agents to allow users to view, edit, and delete the history of the user's data sharing with each.
Users may have expectations of how their data is used, in particular related to data retention, use for other purposes, and sharing. How applications specify how they plan to meet these expectations (application policy), how users express their desires (user policy) and contrstraints on data use may all be related to managing these expectations. A license approach similar to Creative Commons may offer a simple manner to address these requirements.
Issue: This section does not currently contain any requirements. Because retention, secondary use, and sharing are largely out of the control of the APIs, it's not entirely clear that it makes sense to have any API requirements about these aspects. On the other hand, one can envision a requirement that supports the ruleset model, such as:
APIs MUST support a mechanism for users to convey their preferences about retention, secondary use, and sharing to applications in the context of an API interaction.
APIs MUST support a mechanism for applications to convey their policies about retention, secondary use, and sharing to users prior to or during API interactions.
Retention is about to user expectations about how long data they provide will be retained, and whether applications must dispose of collected data after fulfilling the purpose for which it was collected.
Secondary Use is about to user expectations regarding whether applications can use data for purposes other than that for which the data was collected.
Sharing is about user expectations on how data will be shared. Once data have been made available to a requester, the requester is in a position to store and redistribute these data, with or without the user consent. Granularity of data shared is an important aspect of sharing.