The dialogue about what Self-Sovereign Identity (SSI) really is, that was started in the blog "The Path to Self-Sovereign Identity" by Christopher Allen in 2016, has not resulted in a consensus today. While some see the ten principles of SSI that Allen proposed as the definition of SSI, they have been formulated as "a departure point to provoke a discussion about what's truly important".
From the perspective of eSSIF-Lab this "what's truly important" relates to electronically supporting parties as they negotiate and execute (business) transactions. From this perspective, it seems reasonable to have the term SSI refer to all concepts/ideas, architectures, processes and technologies that aim to support (autonomous) parties as they negotiate and execute electronic (business) transactions with one another.
We are aware that others have different ideas about what SSI is. That's ok - it is consistent with our ideas that others are also autonomous, self-sovereign parties. However, it does suggest that it may be worthwhile to postulate the principles that eSSIF-Lab uses to underpin its vision, framework, etc.
The purpose of this section is to provide the eSSIF-Lab principles on which our eSSIF-Lab Framework is based. The are the 'axioms' we use for our reasoning, making choices (e.g. for terminology), etc. We think that making them explicit helps readers to decide whether or not they want to adopt (parts of) the framework.
Freedom of Information (Knowledge)
According to (articles 8-10 of) the European Convention on Human Rights (ECHR), people are free to collect, process, store, and express information (about any entity) in any way the like (making a few exceptions for cases when e.g. security of society is put at risk). This made us realize that
- these rights refer to fundamental, biological capabilities that people are born with, and the ECHR is limiting these capabilities for exceptional cases.
- organizations (governments, enterprises), too, have these capabilities. We generalize this by defining the term party such that it encompasses every entity that has these capabilities. Consequently, this term is a very basic one in our world model.
Parties are autonomous (Self-Sovereign)Parties will use their capabilities as they see fit, as they are autonomous (or: self-sovereign, i.e. sovereign over their own knowledge). While they are free to choose whether or not to comply with laws, rules and regulations, they cannot choose the consequences that other parties will attach to their behavior, as these other parties are autonomous as well. We see that in practice: a person that drives a car on a highway with a speed limit of 55 mph in the desert of Arizona may well decide to go faster than that, depending on his assessment of the consequences it may have (of being caught and fined). In particular, every party is autonomous in the data it collects about other entities, the judgements it makes, the characteristics it attributes, etc.
Identities are not owned - Individuals do not control their identity
'Identity' is a word that is difficult to define. Looking into the world we live in, it seems reasonable to say - and so do we - that 'who I am' is not only defined by how I think about myself (my self image), but also how others think about me, judge me, attribute me characteristics, etc. We take this position not because it has more truth than any other position, but because it is a greater help to us as we try to realize the eSSIF-Lab vision than any other position that we have contemplated.
We use the term partial identity (of an entity), as coined by Pfitzmann and Hansen (2010), to refer to all knowledge that a single party has about that entity (which we also call the 'subject' of that partial identity). Thus, the identity (of some entity, called the subject of the identity) is the union (composition) of all partial identities of that subject.
Since everything that some party thinks about a specific entity belongs to its knowledge, it follows that this partial identity is a subset of the knowledge of that party. And since every party owns its own knowledge, it follows that the partial identity of a subject is owned, and hence controlled, by the party that owns the knowledge of which it is part. Also, when a party issues credentials about some entity, it will use the partial identity that is about that entity and that it owns, to produce the data in that credential (provided that the party won't lie, which of course is not a given since parties are autonomous).
From this, it follows that:
- identities are not owned, because they are composed of partial identities each of which is owned by another party;
- a subject does not control its identity, because it doesn't own it;
- a credential that is (exclusively) about a subject contains a subset of the partial identity that is owned by the issuing party and that has the same subject as the credential - provided that the issuing party doesn't lie.
- for our purposes, the notion of identity is not needed (so its definition is irrelevant for us); all we need is the concept of partial identity.
Identifiers are properties - not things
RFC 3986 (p5) says "An identifier embodies the information required to distinguish what is being identified from all other things within its scope of identification". One can easily see that this implies that an identifier is not something that embodies information, but a property of something that embodies information. This 'something' will or won't have the identification property depending on the context. Still, we use the term identifier to refer to a character string that is being used for identification purposes. For details, see the identification pattern and the identifier pattern.
You do not know what others mean
What a piece of data means to a party is what it decides that it means, simply because this mapping between data and meaning (called semantics) is part of a party's knowledge. This poses a communications problem: a party that wants to understand what another party means by data that the latter authors, must have the prerequisite knowledge that allows it to interpret that data using a sufficiently similar mapping as the author used. For digital data, so-called schema's are said to provide such knowledge, but that is not enough. For example, if a party receives a piece of data that includes a "Alice is a friend of Bob", then even if we know the meaning of being a friend, and that friends must be humans, that doesn't say which real-world person corresponds with 'Alice' or 'Bob'. It is a hard problem for authors of data to provide identifiers that other parties can properly dereference, i.e. use to determine which real world entity the author was actually referring to.