Study Raises Concerns About Vote Count

Edward W. Felten
Director, Center for Information Technology Policy, Princeton University
Monday, November 6, 2006; 2:00 PM

Edward W. Felten , director of the Center for Information Technology Policy at Princeton University, was online Monday, Nov. 6, at 2 p.m. ET to discuss their study on the vulnerabilities of electronic voting machines. Researchers recreated possible security breaches, including the introduction of vote-altering code, that could potentially undermine vote counts. The manufacturer issued a response saying that the machine used had outdated software that is no longer in use. The issue of vote counting and accuracy has been under scrutiny leading up to Tuesday's election.

Read the full paper: Security Analysis of the Diebold AccuVote-TS Voting Machine

VIDEO: Officials Express Confidence in Voting System

The transcript follows.


Edward W. Felten: Welcome, everybody. I see there are several good questions already.


Washington, D.C.: I just voted absentee last week because I knew I had to be on a trip tomorrow. What would happen if the whole idea of automated voting were scrapped and everybody had to vote absentee, on paper? All the ballots would be mailed out, (the IRS mails out our tax stuff so that could be done) You'd complete your ballot and could bring in to the polling place or mail it back in like I did. would this be hopelessly slow to count? Or open to fraud and scams? Seems like vote security is hopelessly compromised now as it is. Also have any of the voting "glitches" discovered so far favored Democrat candidates?

Edward W. Felten: Absentee ballots have their security problems too. There's no way to verify afterward that your vote was counted correctly.

Another problem with absentee ballots is that they don't enforce the secret ballot requirement. You can prove to somebody how you voted (by filling out, sealing and mailing your absentee ballot in their presence), which means that you can sell your vote or you can be coerced to vote a certain way.

Absentee ballots have other advantages, but from a security standpoint that may well be worse than in-person voting.


Carrboro, N.C.: Thank you for doing these chats.

They have been using optical scanning voting machines in my polling places for some time, and it seems like a good solution to me, since the real vote is the paper ballot. I wonder with all the fuss about voting technology, why wouldn't a consensus form around optical scanning as the way to go. Do optical scanning machines have problems too that I don't know about?

Edward W. Felten: Optical scan systems are used widely in the U.S. Done right, they do have advantages. The record is human-readable, for example. But they do suffer some of the same drawbacks as old-fashioned paper ballots: somebody might modify a ballot after it is cast, for example by adding marks to it. And the scanning machinery might malfunction, perhaps in ways that are hard to diagnose.

Still, many people I respect recommend optical-scan systems, especially the ones that scan and count ballots right away in the polling place.


Arlington, Va.: With all the security problems that electronic voting machines seem to bring, the solution is usually said to be to produce a paper ballot that can be checked against. Given this, why not just cut out the middlemen (Diebold) and go with entirely paper ballots to begin with?

Edward W. Felten: Computer-based voting systems do have their advantages. They allow blind people to vote on their own, by plugging headphones into the machine and using an audio interface. They help voters spot mistakes, such as overvotes, where the voter marks more than one candidate in the same race.

Where computers get in trouble is in record-keeping. It's hard to tell what is going on inside a computer. It's hard to ensure that records aren't modified after the fact. This is where paper can come in.

There's a lot to be said for systems that use a mixture of paper and electronics. Today, that means an optical scan system, or a touchscreen system with voter-verified paper audit trail.


New Haven, Conn.: Obviously, the great concern about untrackable electronic voting machines is that elections could be thrown, with no way to know that had happened. In such a case, it strikes me that the strongest circumstantial evidence would be a significant deviation of the tallied results from the exit polls. We've seen that on a national level in 2004; are there any states or congressional districts with electronic voting where there have been significant departures from the expected results? What about Nebraska?

Edward W. Felten: There has been a lot of debate about the discrepancy between exit polls and reported election results in the 2004 presidential race. I think the best explanation, based on the available evidence, is that the exit polls were inaccurate, probably because Kerry voters were more likely than Bush voters to be interviewed by the exit pollsters. Even a small bias in how the pollsters chose whom to approach, or in which voters refused to participate in the polls, would have accounted for the differences we saw.


Washington, D.C.: Mr Felten,

Seeing CNN's piece on voting technology, I was left with the impression that your project at Princeton had examined only Diebold's machines. Did you make any attempt to hack any of the other three companies' machines? Why Diebold? And, finally, why would no Diebold representative accept CNN's offer to appear on the program and allay what appear to be well founded fears about the integrity of the system(s)?

Edward W. Felten: We studied the Diebold system because that's the one we had access to. We would gladly study other machines if we had a chance. My own vote will be cast on a paperless electronic system that we haven't had a chance to examine.

Regarding why Diebold didn't appear on the CNN program, you would have to ask them. They have generally avoided situations where they would have to debate us directly.


Falls Church, Va.: In your response to New Haven, Conn - are you then saying RFK Jr's claims about voter fraud (published in Rolling Stone) are unsubstantiated?

Edward W. Felten: As you suggest, I don't find all of RFK Jr's argument's convincing.

Part of the problem here is that paperless voting systems don't keep the kind of records that would allow us to investigate claims of fraud and figure out what really happened. They electronic records say what they say, and you can either believe them or not. One of the most important goals of our election system is to convince the losing candidate and his supporters that he really did lose. Paperless voting systems tend to do that poorly.


Arlington, Va.: Mr. Felton,

Several weeks ago in the Post, an expert recommended that a nationwide testing and certification center be set up for all electronic voting machines used in the U.S. Seeing as one of the greatest challenges with the machines is their complexity and lack of trained personnel in counties and cities to test and maintain them, might such a center help uncover flaws in the machines, require vendors to build them to a national standard and strengthen voter confidence?

Edward W. Felten: More testing is better, especially if the testers are truly independent. Testing is not the entire answer, though, for two reasons. First, you have to make sure that the machine's setup on Election Day is the same setup that was tested. That can be very difficult, because it's not easy to look at a computer and see how it is set up. Second, some kinds of problems -- notably, security vulnerabilities -- are difficult to detect by testing. A system is secure if there is no clever trick a bad guy might try that can undermine the system's integrity. It's hard to think up and test in advance all of the clever tricks a bad guy might try.


Gaithersburg, Md.: Trust - given the nature of secret balloting, is there a way to build a foolproof system that ensures secret balloting that everyone trusts?

Edward W. Felten: There's no way to build a foolproof system. What we want is a system that is robust enough that it will detect the vast majority of problems so that we can address them.

We're at an interesting point in the history of election technology. We know how to use computers to run elections, but we're just starting to learn how to safeguard computerized voting. Eventually we'll figure out how to build an election system that is more secure and reliable than ever before. I think computers will play some role in that system.


Washington, D.C.: Given that we are not going to have a sudden "change the way everyone votes to the best way (whatever that is)" moment, what do you see as the two or three most important, and most feasible, short term steps that can be taken to improve confidence in elections, starting of course with the Presidential vote in 2008?

Edward W. Felten: The most important safeguard is a voter-verified paper audit trail (often abbreviated as VVPAT). With a VVPAT, the voting machine prints out a paper record that the voter can inspect (usually behind a glass window in the voting machine) and verify that the paper record reflects the voter's intentions. If everything is okay, the voter presses a button to cast the vote and the paper record is kept by election workers as a record of the vote. If there is any dispute afterward, the paper records can be consulted and compared to the electronic records; and a few randomly chosen precincts can be audited routinely after each election to make sure the paper and electronic records match.

Beyond the VVPAT, we can improve election procedures (and follow existing procedures more rigorously) and we can have more independent inspections of voting machines.


Washington, D.C.: Why are there so many conspiracy theorists out there when it comes to these new voting machines? As a voter in the democratic process, we must have faith in our system, or we no longer have a democracy. These people who claim fraud after every election do nothing more than destroy the democratic process. If my vote is changed by a computer or someone else, so be it. If we assume our vote is going to be changed, we might as well not vote at all. Democracy in a way is like a religion, and without faith, there is no democracy. All of these stories and fables of voter fraud and vote switching do nothing but discredit the democratic process!

Edward W. Felten: Certainly distrust can be corrosive. But it's also dangerous to trust a system if it doesn't deserve our trust. For me, the question is whether our voting system deserves our trust. Some parts of the country use systems that seem trustworthy. Others rely too heavily on the security of poorly-defended computer systems. When there is a problem, I think we as citizens should speak up and demand change.

In the words of Ronald Reagan, trust but verify.


Arlington, Va.: Thanks for taking my question.

How easy would it be for a software developer to secretly alter the code in a machine to overcount votes from candidates from one party. Could they put the code inside of some sort of virus that is only activated on election day?

Edward W. Felten: Our research found that the kind of virus you describe is possible on the Diebold system we studied. (For details see On other vendors' systems, we can't say for sure.

This kind of virus is much less dangerous if there is a VVPAT: either the paper record will be wrong and voters will notice this, or the paper record won't match the electronic record.


Madison, Wis.: Why would we ever allow companies like Diebold to dictate to us the terms under which they would provide electronic voting machines (such as contractually prohibiting us to know how their machines operate)? The HBO documentary, whether totally correct in the long run or not, brought me to tears. If the votes aren't counted as cast, what good is our democracy?

Edward W. Felten: I too find it pretty remarkable that we as citizens aren't allowed to know what is going on inside the voting machines. Transparency has always been an organizing principle of our election system -- citizens have been allowed to see how votes are collected and counted.


Washington, D.C.: I would like to see a multiple count/check system. That would mean a paper audit trail produced by the electronic machine that also has a bar code or some kind of marking that is readable by an optical scanner. So every election is counted electronically and then fully verified using optical scanners and random hand counts/checks. Would this be feasible? Would it be cost prohibitive?

Edward W. Felten: There are lots of creative systems like this we should consider. Sometimes the consequences of a new system aren't obvious in advance. For example, if paper ballots come with a barcode or other markings that a voter can't understand, some voters will worry that the barcode can be used to identify them. Can we make a barcode-like system that doesn't raise these concerns? Are these concerns a small price to pay for a system that is otherwise better? Every new design seems to raise these sorts of questions. That's why evoting is so interesting.


Edward W. Felten: I'm afraid our time is up. Thanks to everybody for a stimulating conversation. Please remember to vote tomorrow.


Editor's Note: moderators retain editorial control over Live Online discussions and choose the most relevant questions for guests and hosts; guests and hosts can decline to answer questions. is not responsible for any content posted by third parties.

© 2006 The Washington Post Company