Diebold Inc.'s touch-screen voting system carries a ``high risk of compromise'' by computer hackers and untrained poll workers who could damage the accuracy of election results, an independent report prepared for Maryland has found.
But state officials said the weaknesses can be corrected before its March 2004 primary election, and Maryland is going ahead with statewide installation of the machines.
The report, issued this week, said Diebold's system contains 26 high-risk vulnerabilities that need attention. ``The system, as implemented in policy, procedure and technology, is at high risk of compromise,'' the report said.
The state's board of elections said some of those shortcomings have already been corrected and it laid out a schedule for dealing with the remaining issues. The plan calls for additional training in the coming months for 8,500 election workers and administrators.
Green-based Diebold downplayed the harshness of the report. The company, which has supplied more than 50,000 electronic voting units in Georgia, California, Maryland and other states, said the report didn't find much fault with the equipment or software. Instead, most of the criticism was concerned with how the machines are set up and monitored by elections workers.
``We have never had any doubts that our touch-screen system has provided the utmost in security and accuracy,'' said Mark G. Radke, Diebold's director of voting industry business development.
Some of the report's criticisms, he added, concerned brand-new elections security standards that were adopted by Maryland. Diebold has made several modifications to its software, at no additional cost to Maryland, to meet those standards, he said.
Diebold's contract with Maryland is valued at as much as $55.6 million.
Maryland commissioned the independent study after researchers at Johns Hopkins University in Baltimore released a study over the summer attacking Diebold's AccuVote-TS electronic voting machines.
The Hopkins study concluded that a single voter, using a homemade ``smart card,'' could fraudulently cast hundreds of votes. It also said that poll workers and outsiders also could distort election results by deleting votes, tampering with counts or delaying the start of an election.
The lead researcher in the study, Avi Rubin, said he was ``astonished and deeply disappointed'' to learn that Maryland officials were moving forward with the purchase of Diebold machines.
``State election officials believe all of the vulnerabilities in this system can be corrected by next March. I do not,'' Rubin said in a statement. He added that software is very complex and ``any electronic voting system should be subject to rigorous code review and security engineering practices.''
Some other computer experts also found fault with Maryland's decision.
David Dill, a computer science professor at Stanford University, said Diebold's machines still contain fundamental flaws.
``If you read this report without the spin and know anything about computer security, it says these machines are very dangerous,'' he said.
Kim Alexander, of the non-profit, nonpartisan California Voter Foundation, said relying on poll workers for an accurate election is problematic at best.
The machines and the procedures to run them are complicated. Poll workers are often under trained and underpaid, she said.
``If there's a glitch, it's unlikely it would be detected, and if it is, it's probably going to be too late,'' she said. ``Once the election is over, you can't call them (voters) all back.''
In Ohio, Diebold is one of four vendors approved to sell electronic voting machines to all 88 counties. The others are Hart InterCivic, Sequoia Voting Systems, and industry leader Election Systems & Software.
The state is still conducting security reviews of all the approved machines and a purchase schedule will be set up once the results are in hand, Ohio officials have said.
Counties that choose Diebold will get the brand-new AccuVote-TSX machine instead of the hotly debated AccuVote-TS.