Deployment and Usability of Crypto Credentials (TH5K)
Session Topic: Employment and Usability of Cryptographic Credentials
Convener: Francisco Correlia, Karen Lewison
Notes-taker(s): Karen Lewison
Tags for the session - technology discussed/ideas considered:
Privacy-enhancing technologies (PETs), PKI certificates, TLS, HTTP, cryptographic credentials, privacy, NSTIC, deployment, usability, Idemix, U-Prove
Discussion notes, key understandings, outstanding questions, observations, and, if appropriate to this discussion: action items, next steps:
Privacy-enhancing technologies (PETs), PKI certificates, TLS, HTTP, cryptographic credentials, privacy, NSTIC, deployment, usability, Idemix, U-Prove
Discussion notes, key understandings, outstanding questions, observations, and, if appropriate to this discussion: action items, next steps:
Link to the Power Point presentation which was the basis of the discussion:
http://pomcor.com/documents/Deployment.ppt
Another relevant Power Point presentation for background about privacy-enhanced credentials: http://pomcor.com/documents/ProsAndCons.ppt
Presented a scheme for levels of privacy (levels 0, 1, 2, 3), and the current deployment problem for cryptographic credentials which are needed for privacy above level 0. (Slides 3 and 4)
Three proposed missing technologies needed to implement cryptographic credentials are listed in slide 5--an audience member suggested that a fourth technology, revocation technology for PKI certificates, was also an issue.
For both deployment and usability reasons, it was then proposed that cryptographic credentials should be managed in the browser (slide 6). There was some discussion about the ability to synchronize credentials between browsers on different devices, but most people agreed that it could be done without a lot of difficulty.
Privacy-enhancing cryptographic credentials should be "built into the fabric of the Web" by support by HTTP and TLS; TLS to reduce "overhead" and for credential presentation by the browser (slides 7 and 8). The point was raised that this might be too much for browsers to expect to do; Francisco commented that this scheme was only a suggested architecture. The issuance and presentation protocols for PETs are very similar, so these crypto protocols could both be run within TLS, and PRF can be used to generate a common reference string.
The user experience could be managed by the browser, and include different personas associated with different credentials (slide 9), which would in turn require user education re: persona management. One open question is: Is the user entitled to know the privacy provided by a given credential, and how to explain that to the average user? (slide 10) Alan Karp has been recently involved in usability testing and provided several insights including: the concept that privacy expectations depend on the user's trust relationship with the provider, and that the use of a "private browser mode" might serve as an indication to the user that the privacy level is elevated.
Tony Nadalin added that Microsoft had the same substantial problem with user education with InfoCard.