Problems in E-Voting in Alameda County

To the Alameda County Board of Supervisors, June 27, 2005

Introduction

I am Jerry Berkman, a Berkeley resident, retired after 30 years working at UC Berkeley. I have been a computer programmer since 1969 and earned a computer security certification from the SANS institute in July, 2003.

I urge you to evaluate all possibilities for replacing our current voting system, not just a sole-source contract for a new system with Diebold, for the following reasons:

Is it Legal

The new Diebold system violates a slew of sections of the California Elections Code. The Certification process has also violated a large number of the Secretary of States Procedural Rules for certifying Election Systems. Plus the AVVPAT violates the California standards for AVVPATs. (See pages A1-A3 for details.) Thus even if certified, the County might be sued if it purchases and uses it.

Replacing the TS/Gems 1.18 system with a TSx/Gems 1.22G system would be purchasing a new system; not an upgrade. As such, it is subject to the Alameda County Administrative Code Chapter 4.12.020 and must be put out for competitive bid, rather than a sole source contract.

Reliability

Diebold products have not proven reliable. In the March 2004 primary, thousands of voters in San Diego could not vote because faulty power switches on Diebold encoders prevented the polls from opening on time. Alameda County also had problems with Diebold DREs in that election.

This is not an isolated problem. VotersUnite.org has a 20 page list, "Diebold in the News - A Partial List of Events", http://votersunite.org/info/Dieboldinthenews.pdf which list some newsworthy Diebold failures across the country (in reverse chronological order).

Maryland had major problems with Diebold equipment in the last Presidential Election, 7% of the Diebold DREs failed and 5% were "suspect" because they had too few votes. Plus some of the PC memory cards which stored the vote totals were unreadable. As of March, Diebold had still not figured out what went wrong and had shipped Maryland systems to Houston and Ohio for further testing. And all Diebold equipment in Maryland has been in a "lockdown" mode since the election. See pages A4-A5 for details.

Alameda County Problems, Nov., 2004

The Election Incident Response System, sponsored by a group of non-profits, set up a an incident reporting hot line for Nov. 2, 2004. Bernard Windham compiled an unofficial summaries for several states. The summary for Alameda county is 4 pages long, see pages A6-A9.

Security

Touchscreen computers can not be made fully secure.

The article "E-Voting Security", an article in the IEEE-Computer Society Security and Privacy magazine, page A10, describes how an individual rigged 34 slot machines in Nevada for four years before he was caught, and he didn't even touch the machines.

"Six ways to attack a Touch Screen Computer", page A11, tells how vulnerable touch screen computers are. The Easter egg story is quite interesting.

"Five Myths of invulnerability in touch screen voting systems", page A12, describes more vulnerabilities.

The article "5-27-05: Optical scan system hacked", pages A13-A17, from the BlackBoxVoting.org Forums, describes how both the Diebold DREs and Diebold tabulator were hacked in experiments. One problem of the tabulator is it reads and executes code on the removable memory card.

Many studies have been done (John Hopkins, RABA, SAIC) that show the Diebold systems are not coded with a view to security. The system is based on Windows 2000; a system developed before Microsoft took security seriously. Why Windows 2000 and not Windows 2004?

There have been public demonstrations of how a voter may change the program in a Diebold DRE and one modified memory card can manipulate the tabulator.

Public Trust in Election Results

The public is becoming more and more skeptical that their vote is counted as cast. Reasons include:

On video tape, Congressman Peter King making a comment at a White House function before the election had been finished that, "It's already over. The Election's over. We Won." When Pelosi asks, "How do you know that?" King replies, "It's all over but the counting. And we'll take care of the counting."
In Carteret, North Carolina, votes by more than 4000 early voters were lost in the last election due to a misunderstanding between the vendor and election officials.
Ohio in 2004 and Florida in 2000 were disastrous; there was no respect for the public will.
The polls and actual results differed in Nov. 2004, but only in swing states and often differences were correlated to types of equipment used.
In New Mexico, the certified results show more votes in some precincts than voters.
In Medford, about 27% of all votes cast were not counted due to a programmer error (see pages A18-A19).
The state gives us evaluation reports with part of the security section redacted.
In Georgia, Senator Max Cleland lead substatially in the polls, but lost in the election, just after paperless DREs were installed statewide.
In Nebraska, on ES&S equipement, Senator Hagel pulled an upset in 1996 winning despite the polls; he had been CEO of ES&S.

Public Trust in Diebold

Diebold has especially earned distrust through its actions:

Diebold's past performance in California led to a Secreatary of State investigation and decertification of TSx systems in 2004. Diebold's president was quite arrogant at the hearings. He called the late poll openings in San Diego due to Diebold failures an "inconvenience", while I would call it disenfranchisement. He called the John Hopkins study a home-work assignment.
Maryland's Diebold systems experienced all sorts of problems in the last election (See pages A4-A5).
Diebold's CEO is a Bush "Pioneer" raising $600,00 for Bush/Cheney at one event.
Diebold's CEO said: "I will deliver Ohio for Bush".
Diebold has a history of employees or consultants who are ex-cons; e.g. Jeffrey Dean, a 23-count embezzler who specialized in computer fraud and putting "back-doors" into software (see pages A20-A21 or look in the online version of "Black Box Voting", on pages 83-84.

Optical scan performs better than touch screen

Survey after survey show that optical scan perform better than touch screen on overvotes:

The Cook County (Chicago, Illinois), Clerk's analysis, pages A22-A23. The Cook County study shows differential voting patterns in different demographics in optical scan versus touch screen.
"A growing controversy in West Palm Beach Florida", pages A24-A25

HAVA requires only one disabled accessible voting station per polling site and it can be a Ballot Marking Device instead of a DRE.

Diebold's AVVPAT

It seems designed to fail. The type is too small to see without a special magnifying glass, it is hard or impossible to read the top and bottom lines, it is on thermal (heat sensitive) paper, etc. See pages A26-A27 for pictures and pages A28-A29 for an evaluation of types of paper.

Outsourcing

With Diebold DREs, we can not run elections on our own:

On election day, we need roving teams of Diebold techs to help poll workers, who often reboot the systems, reinstall software, etc. even though they are not licensed or bonded, may be from out of state, and have not sworn to the poll worker's oath.
In many jurisdictions, the registrar's staff can not figure out how to program the ballot definitions and instead contract with Diebold to do that for them.

This is out-sourcing our elections to unknown parties.

The Elimination of Neighborhood Polling Sites

In recent years, an increasing percentage of the population is using absentee ballots. This is due to a number of reasons:

Many people instinctively distrust technology, and would rather vote on paper.
Many have heard of the DRE controversies and vote absentee to avoid DREs.
Snafus at the polling sites, discourage on-site voting (think San Diego in the primary).
The cost of DREs leads to combining precincts in polling places further from where the voter lives.
The need for poll workers with computer skills makes it harder to recruit poll workers.

This is not a good trend. We must return to paper ballots, with the one ballot markup device (but not DRE) per polling site to restore our citizens faith and ability to vote in neighborhood polling places.

I voted absentee last November, by turning in my absentee ballot at my polling site. There was an hour long line there. Why are we paying so much for higher technology if it leads to long lines, and discourages people from going to the polls?

Instant Runoff Voting (IRV)

Oakland, Berkeley, and San Leandro have approved IRV. Diebold in its response to the Alameda County RFP (Request for Purchase) in 2002 said it could do it. Now it says it will cost $1 million and not be ready until 2008. But in response to San Diego's RFP, it says it can do IRV now. Why the discrepancy?

Cost

We currently have a mixed system, optical scan for absentee and provisional ballots and DREs for the polls. Alameda County has under 800 polling sites. We should be able to buy one HAVA compliant ballot marking device plus a precinct level optical scanner for under $3000 per site. This would be $2-$3 million for the county, well within HAVA and Prop 41 funding, and less than the Diebold's cost.

HAVA only covers the initial cost of acquiring the system. Since the DREs are complicated to run, the county must hire DRE technicians, at $1500 each*, to help in the precincts. Also, there is a yearly maintenance fee for the DREs, used or not, of about $350,000*, plus other maintenance and license fees. (* estimates based on unit costs in other states).

The DREs need to be actively cared for between elections. Arapahoe, Colorado failed to keep charging their DREs' batteries between elections and had replace them all for over $100,000.