WASHINGTON --- A national conference of computer security specialists here yesterday concluded with almost universal agreement that new touch-screen voting technology the federal government is spending hundreds of millions of dollars on since the 2000 election debacle may be readily vulnerable to errors and tampering.

"I believe democracy is actually at risk because of ... electronic voting," said David Dill of Stanford University. He told the audience, "Democracy rests on your shoulders. I mean ... it rests on the shoulders of the computer security community."

States are slated for as much as $150 million each in federal money through the Help America Vote Act for new voting technology known as direct recording electronic machines and training of election officials. But a number of states are in the midst of a major controversy and study on the whether the system can be up and running soon enough and whether the system will be open to fraud.

Rep. Marcy Kaptur, D-Ohio, has become one of those pressing for more study into the possibility of fraud in electronic voting and is charging that not enough money authorized under the new law is arriving in states fast enough to train local election workers for next year's presidential election.'

Ohio is so worried about being ready with new technology that some county commissioners want to move the state's primary from March to May.

The research of Aviel Rubin, of Johns Hopkins University's Information Security Institute, has been widely cited as proof that there are major problems because the new law encourages electronic voting. He said yesterday his research found "serious problems" with software for such machines and that the code for one popular machine was even widely available on the Internet.

He said that at least one vendor of the machines and one high-ranking state election official who has bought the equipment tried in vain to get him fired after his research findings became public, even writing to the president of Johns Hopkins. "[Vendors] have a lot at stake," he said.

The machines of Diebold Elections Systems, a prominent maker located in North Canton, Ohio, were the target of Rubin's research and are in use around the country. The firm declined an invitation to appear at the yesterday's symposium.

Diebold, however, has counter-charged that Rubin used the wrong software, hardware and environment to conduct his test and failed to take into account the use of poll watchers who could prevent so-called smart cards from being used by more than one voter. Rubin said Diebold's charges about his research were not true.

Douglas Jones of the University of Iowa said he also found flaws in Diebold's machines years before Mr. Rubin published his results and that they had not been corrected.

The security symposium, sponsored by the Advanced Computing Systems Association, featured a panel of academics, a state election official and one vendor.

The moderator, Dan Wallach of Rice University, joked that he was concerned enough about the controversial subject that his job was to prevent "blood from flowing off the stage" and that if anyone got too boisterous, "I will give your name to the National Security Agency and have you declared a terrorist."

One basic problem widely discussed at the symposium was the difficulty of voter verification and how to get an audit trail. There are many concerns that local officials could skew the machines to record votes falsely.

Sanford Morganstein, a vendor who showed up and was widely cheered for his "courage," represented Populex.

"We welcome criticism. If we've got a thin skin, we shouldn't be in this business," he said. He said his company was asked to get into the business by the state of Illinois because of the history of voting "fraud in Chicago."

He said his company's system used touch-screen technology that prints out the voter's selection on a card with a barcode that the voter must drop into a ballot box to have counted.

"Can it be fooled? Yes. Can it be better? Yes," he said.