PDS Architecture

{{#eclipseproject:technology.higgins|eclipse_custom_style.css}}

A PDS provides a central point of control for information about a person, their lives, friends, interests, affiliations and so on. The information stored locally within a person's own PDS is entirely controlled by them. The PDS may be operated by an organization, but the notion is that that organization works as an agent of the individual and does as the individual dictates.

A key point to understand is that in many cases what's stored in the person's PDS is a pointer to data stored elsewhere. Only a portion of a person's data is physically located on their own PDS. The person typically has a bundle of rights in these external data, but far less than complete control. In the diagram below you see one of these pointers pointing to a PDS operated by, say, AT&T. The data in the person's AT&T PDS would contain information related to their account at AT&T and perhaps usage history, current closest cell tower, etc.

What's not shown below is that a person might have a pointer into a data object stored in a friend's PDS. These pointers, taken together, form the social graph that is physically distributed across the PDSes.

A PDS does centralize control by providing the person with a dashboard or control panel, but it does not centralize storage. Quite the oppostie. At the heart of the PDS concept is the idea of a distributed architecture. Theoretically every person could run their own PDS, perhaps in the cloud.

In the diagram above we show a person's PDS and some apps and service providers acting as relying parties across the top. Across the bottom we show some external "data sources" that act as sources of a person's attributes. Note that a given party may well play BOTH the role of being a relying party AND a data source. In the former role they act as a consumer of attributes and in the latter role they act as an authority for attributes.

Among many benefits, the PDS architecture tends to shift much more control to the individual over their own data. On the other hand the new architecture also introduces a number of new challenges related to governance, access control, interchange standards, authorization, auditing and so on. These challenges are being addressed by the wider ecosystem within which Higgins plays its role as a provider of a PDS codebase.

We all play different roles and share different sub-sets of our social graph and attributes depending on who we're interacting with. That's what humans do. For this reason a single person is represented as a set of partial identities that are used in different contexts. Some of these are generic enough that we can curate them as a handful of defined "personas"--that way we can reuse them in a specific context (e.g. buying something at a new eCommerce site) without having to tediously repeat ourselves. A hint at these different personas and contexts is suggested in the diagram above.

The vision many people share is an internetwork of PDSes that can freely exchange information of a wide variety of types. Some will be based on Higgins code, some based on other technologies entirely. However, until a single protocol emerges as the universal standard, we anticipate implementing whatever protocols get traction in the wild.

Architectural Overview

The diagram below adds a few more details to the overview above. For simplicity it only shows a single PDS and omits the "AT&T" PDS shown above. The PDS is shown below as comprised of two distinct services: PDS and PDA that are described further below.

PDS

The PDS is shown in the middle of the diagram above. Information from a wide variety of data sources such as the social network, telco and health data sources are virtually integrated by the PDS and presented in a "dashboard" application in a browser (via the PDA) or in desktop and mobile clients. The PDS gives you control over your own information by allowing you to share selected subsets of it with other people and organizations that you trust.

• Gives you control over your information stored in hundreds of external silos as well data stored internally
• Provides a personal data abstraction layer mapping internal and external data sources into a consistent data model based around notions of personas
• Manages a set of your personas (e.g. Work, Home & Friends, Citizen, Health, Anonymous)
• Provides an encrypted "lock box" in the cloud such that certain internally stored data in the PDS (e.g. your persona definitions) cannot be read by the PDS's operator
• Backs up personal data stored on your desktop and mobile devices
• Synchronizes personal data to other devices and computers owned by the person using a variety of network protocols.
• Manages access by external apps (aka service providers) to your data that is stored locally as well data stored in PDSes managed by external organizations
• Links information from your personas to accounts (profiles) that you have at services providers, websites, social networking sites, etc. and over which you share joint control and rights
• Links information from your personas with the personas of your friend's and colleague's PDSes

Although in principle you only need a single PDS to work on your behalf, we expect that until the time comes that most people fully trust the architecture's security and privacy characteristics, a significant number will choose to have more than one PDS account. Doing so allows them not to have to trust that the PDS architecture is capable of fully isolating the private, work, health-related, and social dimensions of their lives.

PDA

Shown at the middle left of the diagram above:

• Provides a web client for your PDS (as an alternative to native PDS clients on desktop and mobile devices)
• Provides a run-time environment for apps that run within the PDA itself
• Implements an API that allows you to be discoverable by people and organizations that meet criteria you specify
• Decrypts data from your PDS (using a locally stored key) to allow it to be managed in the PDA's "dashboard" UI. Attribute data stored locally on the PDA are encrypted by the PDS Client using an internally managed key prior to transmission to the PDS. Thus data attribute values on the PDS are blinded from the service operator offering/hosting the PDS.
• Implements two factor authentication

Native PDS Clients

As shown at the top of the diagram above, we are developing native Windows, Mac and mobile clients for the PDS. These clients have two advantages over the web-based PDA. First, data stored on these devices is entirely under your control without the need to rely on third party hosted services. Second, the client is closely integrated with the browser and other local apps. This allows the client to capture information about you as you browse and can augment your web experience through web augmentation (overlaying context-specific information within your browser) as well as through automatic form filling (e.g. filling in your passwords).

Representing a Person

A natural person is represented in a PDS as a set of containers called contexts each of which holds a partial digital identity called a persona. Each persona instance has a set of attributes and values. Thus one individual (natural person, data subject) is represented as multiple personas each in its own context-container.

The data in these contexts adheres to the Higgins Persona Data Model 2.0, which can be used for storing arbitrary (identity and social networking) data. UDI references are used for representing links between contexts, both inside the Personal Data Store 2.0 and to external data stores.

Components

• PDS 2.0: Personal Data Store core service
• PDA: We have not even started developing this component for Higgins 2.0. We developed something similar in Cloud Selector from Higgins 1.1. Similar in that it was a pure web app and that it was a "client" of the core PDS.
• PDS Client 2.0: a library used to access the Personal Data Store 2.0. It is incorporated into the PDS agent as well as PC and mobile PDS clients.
• Authentication Service 2.0: is an OAuth web service that authenticates PDS users and returns an access token that is relied on by the PDS Agent and the PDS Vault.

Data Models

Data models used in Higgins code and services:

• Persona Data Model 2.0
• Higgins Data Model 2.0
• Context Data Model 2.0

IdAS 

The IdAS solution is a testbed for exercising the IdAS Java framework.

• Higgins 1.1: See Higgins 1.1 Plan#IdAS_Solution_1.1
• Higgins 1.0: IdAS Solution 1.0: a basic configuration of the Identity Attribute Service 1.0 (IdAS). IdAS is a java framework that provides a common interface to identity, profile, and relationship data from external data sources (e.g. websites, databases, directories).

XDI4J

XDI4J is a java library for working with XDI.

• Higgins 1.1: XDI4j 1.1