Sep 15
Sam's notes...
So I wasn't able to be as productive reading as I'd hoped this past weekend, due to some health stuff and putting some of my time into coding for AI. I still got the last three trust papers I really wanted to get read done, they're 'burrowsLogicAuthentication', 'abadi_logic_semantics', and 'TrustModelAndEval' (in Drive/Identity Auth/Trust Papers/Read). The first two were associated, the former a very commonly cited paper building a formal logic for describing authentication relations (it considered trustworthy actors and hostile attackers, not untrustworthy actors, however) and applying it to a variety of protocols (Kerebos the only one commonly still used, not counting an early X.509 based one that's a simple version of current models), the latter an attempt to give a deeper semantic meaning to some of the handwaving in the first. Useful, but the second still waved its hands at the concept of "honesty". Still, I like their logic, and I can see why they've been influential papers, dating back to the late eighties. The last paper was an attempt to build a trust propogation mechanism using a algebraic structure called a semiring. Their system isn't horrible, although it doesn't include any notion of evaluating trust at different levels for different actions and leaves the act of actually quantifying explicit trust in the air, which are the two major issues. Again, maybe in a very limited case some utility, but I'm not buying it generally.
In addition, I did a bit more thinking about my final plan component. I think I've got a general structure for how the formal language would work, which I'll run by you when we meet tomorrow. The main thing I've realized needs to change is that the "telescoping" effect needs to be built off of the abstract description of the processes, and not determined hierarchically off the various high-level actors. The latter leads to too many interlocking routes and will be a lot "uglier" and harder to use than the mechanism I'm imagining now.
For next week, my goals are the IBE/ABE papers I have set aside. Since I need to do some writing for Matt associated with them (which will also be useful down the line for the identity auth paper), those six or so papers should be a good haul, especially with their density. I'm thinking another 3-4 weeks of reading (IBE next week, then a week for PKI, a week for non-certified means (web of trust and passwords) and a week for some other general crypto papers), then start writing a lot, while still periodically looping back to read more papers.
