A firewall for democracy (touch-screen voting machines denied certification, 10% errors)

LA Times ^ | August 4, 2005 | Andrew Gumbel

[neverdem]

Two California secretaries of state have refused to certify a touch-screen voting machine, and for good reason.

BRUCE MCPHERSON, California's secretary of state, has just performed an invaluable service for the voters. Only a few months into the job, he had been under intense pressure to certify the latest electronic touch-screen voting machine manufactured by Diebold Election Systems, which is supposed to help California counties meet a federally mandated January deadline for the overhaul of their election equipment. But instead of rolling over, his office conducted exhaustive tests on the Diebold TSx, discovered that it had a 10% error rate — worse than the reviled punch-card machines used in Florida in 2000 — and sent the company back to the drawing board.

This is hugely important for two reasons.

First, because companies such as Diebold have been operating far too long without adequate public oversight, while charging taxpayers tens of millions of dollars for equipment that some of the country's best computer scientists have found is sloppily programmed, dangerously insecure and virtually unable to be audited for fraud or error.

The TSx, which is fitted with an independent paper trail to make recounts possible, was waved through the hazily regulated federal certification process and was deemed to have passed muster in Utah, Ohio and a handful of other states. None appear to have noticed the paper-jamming and screen-freeze problems picked up in the California tests.

McPherson's decision is also significant because the same line of Diebold machines has now been refused certification by California secretaries of state from both major parties. McPherson is a Republican, and his predecessor, Kevin Shelley, who cracked down on Diebold last year after the company admitted shortcomings that included the use of uncertified software, is a Democrat. That shows the vulnerabilities of the electoral process...

[inkling]

Weren't the Dems the ones that demanded touch-screen balloting after Florida in 2000?

[Amerigomag]

It will soon become both prudent and wise to vote absentee so that a voting record is preserved.

[Carry_Okie]

Here's a pertinent piece I wrote in December 2002.

Ballot Transparency to Eliminate Fraudulent Counts

Voters have read and seen all sorts of assurances that the new touch-screen balloting systems are fool proof, tamper proof, and nothing to worry about. Many, including those who are familiar with the technology, are not at all reassured.

The concerns are on two levels. First, from the perspective of those not familiar with the technology, it is a device whose inner workings and inherent security they cannot possibly understand. If they can't understand it, how can they be assured that it is honest? Second, those who DO understand signal processing, software, and communications technology know that is far too easy to defraud the system in a way that would be irreversible and undetected. Either way, touch-screens are a loser.

Now, as users of ATMs, cell phones, the Internet, and other electronic media, it might at first seem a little strange that so many people have such concerns. Upon further consideration however, the key distinctions between voting and a service handling mere money become obvious:

• Customers have a choice of banking vendors. Citizens don’t have a choice of governments.
• There is a major difference between mere financial assets at risk, and a risk to individual liberty.

Governments are monopolies. One can go down the street to another bank and take the offending bank to court. An evil government can land you in prison (or worse) because they ARE the court. The stakes associated with voter fraud are far higher than with an ATM and so is the temptation to defraud the system.

Necessary and Sufficient

So, given that we are still smarting over hanging chads, what are the alternatives? Let’s begin to answer that question by looking at the requirements.

1. The system has to be simple and familiar to the voter.
2. There must be NO SOFTWARE involved, because it is too easy to change.
3. The system must be capable of completely manual operation.
4. The count must be capable of being validated by all parties involved and each count must be separate and distinct.
5. There must be no possibility to count a ballot twice or "lose" counts along the way.

Electronic sensors and interlocks are permissible as long as they can be duplicated manually.

Here is my proposal for a system that meets these requirements:

At the Polling Place

1. Ballot boxes are preprinted, serialized and tracked by a physical chain-of-custody document.
2. The box must be destroyed to be opened.
3. The box is locked under a ballot receiving machine.
4. The ballot receiving machine at the polling place reads the box number and records it on the ballot in Scantron form on the back side (fill in the dots). Note that one could do the same manually under observation.
5. The voter completes the standard optical ballot and delivers it to the receiving machine.
6. The machine prints the box number on the back of every ballot it accepts with a Scantron dot pattern. This too can be both read and performed manually. Then a dry film coating (basically an adhesive or heat activated tape) is applied to the ballot on the way into the sealed ballot box.
7. The coating is transparent but reveals a "watermark" when exposed to UV light. The ballot is now tamperproof.
8. The receiving machine totals the number of ballots in every box. The total is read manually and a receipt is delivered to each political party and candidate detailing the box numbers, precincts, and tally of ballots in every uniquely identified box.
9. Representatives of all Parties check the box tallies before the boxes leave the polling place.
10. If they agree on its accuracy, they record the ballot tally on the box using Scantron dots, initial it, and put a similar dry film over the number.

Note that the Scantron pattern is the perfect bridge between human and machine. It is readable by people for manual counting but does not require an optical character reading machine that needs cameras or software.

Both parties thus know the EXACT number of ballots cast in every precinct and in every box. Every box is signed. All parties can thus run check sums at the processing centers and verify the chain-of-custody.

At the Ballot Counting Center

1. The total of the ballots on the box is read by the counting machine. It would be very similar to the existing optical reader and might only require very minor modifications.
2. The counting machine reads the box code for precinct and ballot count or accepts that data input from a keypad read off the box by at least two witnesses with keys. The machine will not count the ballots without the UV visible watermark on the ballot over the votes AND matching precinct codes on the box and the ballot.
3. The machine halts and will not display the vote totals if the number of ballots recorded on the box and the number it counts do not match.
4. The ballots leave the counting machine get a NEW ballot box. Counted ballots are stamped again with output box number, recoated, and then deposited into the new sealed ballot box.
5. The new coating was applied in case of a recount, thus each ballot maintains a recount history.

[Mr. K]

figures... LA Times link pops up an ad for the ACLU

[GOPologist]

See if there is any logic to my reasoning?

There are 10% errors; therefore, only 10% of the Dems' ilegal votes will be discovered.

Yes, no, or am I a "ninny?"

[prophetic]

mega bump

[lepton]

But instead of rolling over, his office conducted exhaustive tests on the Diebold TSx, discovered that it had a 10% error rate — worse than the reviled punch-card machines used in Florida in 2000

Of course the "punch-card issue" was fabricated to begin with.

[Extremely Extreme Extremist]

LA Times link pops up an ad for the ACLU

LOLOLOL!

[lepton]

I do note that no where in the article does it define what a "10% error rate" is.

Does that mean that 10% of the time it tells you you voted for candidate A, and applies it to another candidate? ...Or does it mean that 5% of the time you have to push four times to get it to acknowledge?

[ccmay]

Britain manages to get speedy, accurate eection results with nothing more than tick-boxes on paper ballots.

This electronic voting business is a solution in search of a problem. I'd rather wait a few more hours on election night, than depend on a bunch of hippie computer programmers writing proprietary code.

-ccm

[Political Junkie Too]

In California, they used it as an excuse to delay the recall from October 2003 to the Democrat primary in March of 2004.

They argued that the punchcard system was discriminatory to minorities because they couldn't figure it out, and we should wait until March of 2004 for the touchscreen machines. In reality, they wanted to move the election away from a date recall-motivated voters would turn out (who favored recall), to a date that Democrat primary voters would turn out (who favored Davis).

The California State Supreme Court rejected the argument, and then the Secretary of State refused to certify the touchscreen machines.

-PJ

[kingu]

Nahh, they count a 'non-vote' as an error. Someone using a computer feels less pressure to vote in every race and it is easier to skip issues. Therefore it is defective because people are using it as intended.

[snowsislander]

Never mind that several high-profile studies — such as the Florida commission report on the 2000 election — have concluded that optically scanned paper ballots, not touch-screens, are the safest and most reliable technology available.

Yes --- I also prefer the simple optical scan ballots, the kind that can be read as easily by humans as by machines. I don't like the ballotless electronic touch machines.

[expatpat]

IMO, the best system is the optically-read paper ballot.

[supercat]

IMO, the best system is the optically-read paper ballot.

I happen to like the machine-produced paper ballot myself, but recognize its benefits may not justify the added expense (let voters make their selections via touchscreen or other medium, then have a ballot printed per the voter's specifications; this avoids any possibility of mismarking and also makes practical the inclusion of check-digits and other features to prevent alteration).

[Caesar Soze]

Nahh, they count a 'non-vote' as an error. Someone using a computer feels less pressure to vote in every race and it is easier to skip issues. Therefore it is defective because people are using it as intended.

If correct, that answers a question that had been bothering me ever since seeing the headline. At the risk of sounding like a technophile, computers don't make errors. Not with simple tasks like recording lists or counting things. Every second my computer performs thousands or millions of times as much work as a Diebold machine would during an entire election, and does so with no detectable errors. (Well, I can make the video card driver crash, but that's a special case.) I can, and do, handle theoretically unbreakable encryption, high-resolution protein folding simulations, DVD quality video and audio, and freeping, often all at the same time. The idea that a machine could fail to record little check marks next to someone's name and ++ a counter ten percent of the time is incredible. Not even government contractors could screw that one up.

[MirrorField]

Machine-produced and machine-readable ballots are IMHO also the way to go, but few issues should be met:
1) Open Source Software. This should be mandatory for any computer-assisted voting system to keep both vote-counters and manufacturers honest. Otherwise we're better off with old-style paper and box ballots.
2) WORM (Write Once, Read Many times) audit trail. Paper has served historically, but this is doable by other technologies and will probably change. But not today....
3) All regular requirements for paper voting (ie. vote anonymity, fraud resistance, etc.) must be fulfilled or exceeded.

[THEUPMAN]

At the risk of sounding like a technophile, computers don't make errors. Not with simple tasks like recording lists or counting things

The problems not error .... it's fraud. The numbers can be manipulated from the keyboard, with no log file recorded.
It seems like it's become one of those things that both sides believe they can utilise ... so everyone will just "overlook" that little aspect. Because the ones in controll of the keyboard .. well .. surly we can trust them . Right?