Re:WikiLeaks Posts Mysterious ‘Insurance’ File
by derefr on Saturday July 31, 2010 @12:08AM
(#1563206)
But who decides what "has to" be exposed, and what doesn't? The problem is that there is no third party at all trustworthy enough to bear the responsibility of "filtering" through all the world's dirty laundry to just find the "morally reprehensible" parts. So, there must be some (transparent) strategy set in advance: they either release nothing (most news agencies), everything (Wikileaks), or use some other pre-defined filter that can be proven to function (i.e. they cover everything, but then push their stories to a proxy server, with a transparent and inspectable program designed to only publish the stories that match some algorithm.)
[ Parent
| on HN ]
Re:WikiLeaks Posts Mysterious ‘Insurance’ File
by sgift on Saturday July 31, 2010 @07:09AM
(#1563564)
The filter is already build into WL (and similar organizations): They can only publish information they get from informants. The informants are the filter.
Re:WikiLeaks Posts Mysterious ‘Insurance’ File
by thwarted on Saturday July 31, 2010 @02:53AM
(#1563356)
How do you determine if the transparent and inspectable filtering program is operating correctly if you only have access to the output and none of the inputs? You'd need all of the inputs, and at that point, you just have access to all the information anyway, and wouldn't need the program to do the filtering.
Re:WikiLeaks Posts Mysterious ‘Insurance’ File
by derefr on Saturday July 31, 2010 @03:06AM
(#1563374)
Reliable computing isn't just for NASA; it also works when an individual codebase may have been designed to "malfunction." Instead of a single machine, you contract several independent, mutually untrustworthy parties to implement (open, inspectable) machines, with an architecture that takes votes from all machines into account. You could then spit out a log of (hash of input, vote) pairs from each machine to see that the voting architecture is also operating correctly.
Re:WikiLeaks Posts Mysterious ‘Insurance’ File
by thwarted on Sunday August 01, 2010 @11:08AM
(#1565357)
You'd need at least one trustworthy party because you can not trust the mutually untrustworthy parties to not give similarly faked logs.
If the goal is secrecy, increasing the number of parties that have access to the inputs undermines that, especially if they are untrustworthy, even if they have "contracts".
While the solution you've outlined may be mathematically sound, I don't think it is very pragmatic.