Anonymity and trust in the electronic world
Privacy has never been an explicit goal of authorization mechanisms. The traditional approach to authorisation relies on strong authentication of a stable identity using long term credentials. Audit is then linked to authorization via the same identity. Such an approach compels users to enter into a trust relationship with large parts of the system infrastructure, including entities in remote domains. In this dissertation we advance the view that this type of compulsive trust relationship is unnecessary and can have undesirable consequences. We examine in some detail the consequences which such undesirable trust relationships can have on individual privacy, and investigate the extent to which taking a unified approach to trust and anonymity can actually provide useful leverage to address threats to privacy without compromising the principal goals of authentication and audit. We conclude that many applications would benefit from mechanisms which enabled them to make authorization decisions without using long-term credentials. We next propose specific mechanisms to achieve this, introducing a novel notion of a short-lived electronic identity, which we call a surrogate. This approach allows a localisation of trust and entities are not compelled to transitively trust other entities in remote domains. In particular, resolution of stable identities needs only ever to be done locally to the entity named. Our surrogates allow delegation, enable role-based access control policies to be enforced across multiple domains, and permit the use of non-anonymous payment mechanisms, all without compromising the privacy of a user. The localisation of trust resulting from the approach proposed in this dissertation also has the potential to allow clients to control the risks to which they are exposed by bearing the cost of relevant countermeasures themselves, rather than forcing clients to trust the system infrastructure to protect them and to bear an equal share of the cost of all countermeasures whether or not effective for them. This consideration means that our surrogate-based approach and mechanisms are of interest even in Kerberos-like scenarios where anonymity is not a requirement, but the remote authentication mechanism is untrustworthy.