Reply #222: Endless variations on a theme [View All]

[dlaliberte]

Hi skids... I am a software developer too, willing to help where I can. You seem to have most of this issue well in hand, but the endless possibilities leave plenty of room for more ideas.

A couple thoughts on the idea of a program that can delete itself, or at least the portions of itself that do the vote flipping (or whatever the nasty thing is designed to do). Yes, it could do this, but it could only do it once. The original program would have had the code which does the vote modification as well as the ability to delete itself and both of these components must be deleted in order to leave no traces.

But there would very likely be traces on the old version on the hard disk unless it overwrote those parts of the disk. This just seems unlikely given the complexity. It might be easiest just to self-destruct by overwriting the whole disk rather than appear completely innocent. This would be suspicious too.

Another way to make changes to code that would not be visible afterwards is to load it into the runtime memory, and thus never write it to disk. This could happen by network access or perhaps by typing into a console or loading an external memory device such as a floppy drive (not that I think these evoting machines have a floppy drive). Once the machine is power-cycled, the evidence is gone.

Regarding the user-visible errors, it seems likely that the modification was very limited in what it could do. Even though the error of flipping the voters' selection was noticed by some people, not all people will notice it, and such an error all by itself might account for most of the surplus votes for Bush. Also, we have heard that poll works were telling voters that the summary/confirmation page that showed the flipped vote was wrong and go ahead and accept the vote.

Some people backed up and retried several times before the correct vote showed up in the summary. It is interesting to speculate how this non-deterministic behavior could result. Why would it not do the same thing every time? If it decides (based on the current tally) that it will do the flip, it could use a random function to reverse that decision with probability of say 1/10, or it could do the flip no more than 10 times in a row.

Note that there is no possibility of recounting such a flipped vote to correct the problem because people really did record a vote that they did not intend. Only statistical evidence (e.g. exit polls, or conflicts with other votes on the ballot) would suggest that people did not intend to vote this way.

I think a user-visible error like this is unlikely to be due to some kind of innocent alignment error, or even a subtle user interface error, and thus it would show up in the code if it were examined. So why have we not been able to examine these machines? They are supposedly protected from such examination because the code is proprietary. Well, if they believe this is enough protection, then there is no need for self-deleting code that covers its tracks. Eventually, they need to destroy the evidence before the courts can get around to deciding there is no way to prove fraud in this case without examining the machines.

The story about the 21 evoting machines in Florida that were taken away (to a warehouse?) is very suspicious making you wonder why and what they are doing with them. A memory purge would be very likely.