Earlier this week, we noted that the press was a little bit hasty in claiming that the e-voting machines in Sarasota County, Florida had no problems. That wasn't what the actual study found. It simply found no evidence of problems in the source code -- which is quite different. Ed Felten continues to examine the situation and details why the lost votes seem consistent with a typical computer bug in the software. He's not arguing that it definitely was a glitch in the code, but simply that there's enough evidence that you shouldn't rule it out. His reasoning is that similar bugs have been found in other e-voting software, the undercount behavior is consistent with a common type of computer bug and that the study report from last week would have been unlikely to find that bug. However, perhaps the most interesting point is his pointer to two studies of the e-voting machines that found that machines that showed more undervotes tended to have been treated slightly differently. Machines that were set up a certain way and which were cleared and tested right before election day tended to have higher incidents of undervotes -- suggesting that the activities in setting them up may have triggered a memory error or buffer overrun, which could explain the errors. It's this type of analysis that suggest that a more thorough examination of the machines and the software used would be quite useful -- rather than a very limited test of just the source code. There's no compelling reason why further test shouldn't be allowed.
If you liked this post, you may also be interested in...
- Crowdfunded Prize For Open Source Jailbreaking iOS7 To Improve Accessibility
- Advisory Panel Offers Suggestions To Strengthen US Cybersecurity, But Is The Government Capable Of Change?
- ACLU Calls For Ban On Nonlethal Weapons In Schools After Tased Student Ends Up In Coma
- Lightning Strikes Twice: Wannabe Murderer Butt-Dials His Almost-Victim
- Companies Developing Crowd Analysis Programs To Detect 'Abnormalities' In Behavior And Match Faces Against Giant Databases