Implementation of Bram's trust metric

[ciphergoth]

I've implemented Bram's probablistic trust metric: see this Advogato diary entry:

http://www.advogato.org/person/ciphergoth/diary.html?start=5

Update: - just got mail from Bram indicating that this isn't the metric he wanted to propose. More to follow.

Comments

[kaet]

What are the colours in the .dot file? Link weights? Or are they just prettiness? Or have I not read carefully enough the Advogato setup? Also what's graphviz. It looks like .dot's are easy to parse but if its a pretty graph viewer, that would be cool.

I'm interested because my job used to be developing graph algorithms for network resilience for a telco, so there's a reasonable chance I might be able to add something here whereas much of this stuff is crypto which I don't really do at research level.

[ciphergoth]

You can rate people on Advogato as Observer, Apprentice, Journeyer or Master - the link colours indicate which. The Perl source code gives the map.

graphviz is exactly as you say a pretty graph viewer.

[kaet]

(graphviz) Oh, that's cool. I'm sure that wasn't around last time I looked

[kaet]

I think this model is right. I don't know whether the failures should be node representing "bad egg" probability, or arc representing "bad call" probability, but anyway, the first thing you tend to do is to transform node failures into arc-failures and hypothesise perfect nodes, so you can combine the two.

[fwiw, for the people reading this I'm not teaching to suck eggs, the algorithm for transforming a graph with unreliable nodes and arcs into one with just unreliable arcs is to turn each node into two nodes linked by a single directed arc. On the node at the source of the arc you put all the incoming arcs to the original node (fan-in), and at the node at the destination of the new arc you put all the outgoing arcs (fan-out). Then node failure is represented by the failure of the connecting arc].