Approaching Trust Architecture on the Web

Athens.
Finally for this trust and security panel at WebSci '09 we move on to Pavlos Spirakis. He begins by noting that the Web was initially invented as a tool for researchers who trusted one another implicitly, so no security was built in from the beginning. This is no longer the case, of course, and so theer is a need fro privacy protection and for the protection of the freedom of its users. Layers of trust must therefore be engineered into Web technology.

Trust, however, is the belief of one party in another, and distrust is not simply the lack of trust, but the belief that the other partner explicitly cannot be trusted. Such distrust is an important notion in its own right, then. Trust is a relation between two parties, then, is context-specific, and exists in varying degrees of strength. This means that trust is measurable, Pavlos suggests, and that it can be expressed mathematically (algorithms such as Google's PageRank build on this).

Any specific trust model responds to a particular form of threat - digital trust (the need to validate an entity), transitive trust (trust transmitted through other parties), and assumptive trust (where there is no mandatory or explicit validation of credentials - a 'take it or leave it' approach as embodied in PGP encryption). Again, these can be expressed mathematically.

But how are beliefs represented on the Web? Pavlos points to the rise of the Semantic Web as one model - its Resource Description Framework allows users to define data relations. On this basis, we may evaluate and monitor explicit trust relations on the Web. Additionally, we may use reputation, monitoring, context,and content as input for such trust calculations, too. Trust services, then, implement such approaches in order to deliver trust and confidence in Web transactions.

In theory, this may be further developed into trust architectures on the Web, to provide an integrated view of trust on the Web. Such architectures must be open, should not exclude providers which have not themselves been evaluated yet, should try to use all relevant information, should allow users to formulate subjective and task-specific policies, shoud be able to justify their decisions to the user, and should be light-weight.

Can such systens be created? There are significant challenges to overcome on the Web - implicit data needs to be extracted, the trustworthiness of anonymous entities must be evaluated, the role of conflicting viewpoints must be understood, the Web is dynamic and constantly changes, and situation awareness must be built in.

Can such architectures appear, then? Pavlos believes no: there are so many requirements and conflicting demands that any system will be inconsistent - no total control mechanisms exist. But on the other hand we may try to approximate the perfect architecture, and by extending of the Sematic Web approach we may achieve at least some trust architecture elements. Modern cryptography also plays a role here, by protecting sensitive information, respecting privacy, and reducing overheads.

There is much research ahead. The human user must be placed in the centre here, rather than the system, and the approach must be as interdisciplinary as Web science itself. There must be an open call for participation by a wide range of researchers.

Technorati : WebSci '09, identity, policy, privacy, security, trust
Del.icio.us : WebSci '09, identity, policy, privacy, security, trust