SANTA CLARA, Calif. - Twenty-five years ago, Steven T. Kirsch built a better mouse.
Now he believes he has found a way to create a better trap - for spam, not mice - if he has enough time to finish his project.
An MIT-trained engineer, Kirsch was frustrated by the quality of the first computer mice in 1982, so he set out to improve them by incorporating an optical sensor.
Since then he has started four companies, all based on his frustrations with existing products or services. He has made forays into word processing document design, accelerating the Web, and in 1997 Infoseek, his search engine company, was the third-ranking company in Web search.
In many ways Kirsch, who is 50, has come to exemplify what distinguishes Silicon Valley - a blend of engineering skills with persistent entrepreneurship.
Along the way he has amassed a personal fortune of about $230 million, a success that has permitted him and his wife to become significant philanthropists in Silicon Valley by contributing more than $75 million to the United Way campaign and other causes through his foundation.
Recently he has taken on the challenge of e-mail spam. This year he founded Abaca, a company with a new approach in the crowded market for stopping junk electronic mail.
Abaca says it can filter out 99 percent of all spam, and supports the claim with a money-back guarantee.
According to the result of an independent survey in February by Opus One, a computer industry consulting firm in Tucson, Ariz., that would be significantly better than the results of six leading spam blockers.
Abaca has taken on a new urgency for Kirsch: During the summer, he was discovered to have a rare form of blood cancer, Waldenstrom's macroglobulinemia, that is found in about 1,500 Americans every year and is considered incurable, although it can be managed beyond the five- to seven-year longevity that new patients are usually told to expect.
So far he has shown no effects from the disease, and he said he is intent on applying his engineer's approach to the problem.
"This is harder on my wife than it is on me," he said during a recent interview. "I just look at it as a problem. Here's a problem and you have four years to solve it or you don't get to solve any more problems."
The most visible change that he has made as a result of his cancer is the recent decision to change the financing direction of the Steven and Michele Kirsch Foundation, which until late last month had focused on a wide range of community philanthropic goals.
At the end of October, however, the foundation announced that in the future it would focus its financing on research associated with his cancer, which because of its rare nature receives almost no federal money.
For the moment, between weekly visits to the Stanford Medical Center, Kirsch is continuing to put much of his time into convincing the world that he has stumbled on a better way to block spam.
The approach underlying the Abaca technique is the recognition that the ratio of spam to legitimate e-mail is individually unique. It is also a singular identifier that a spammer cannot manipulate easily. By assessing the combined reputations of the recipients of any individual message, the Abaca system determines the "spaminess" of a particular message.
Kirsch asserts this provides a high degree of accuracy in deciding whether the message is spam.
Unlike most of its competitors, he said, Abaca's technology does not require a training period, is language independent, and is faster than many competitors because it does not scan the entire contents of a message to determine whether it is spam.
Kirsch has invested about $5 million in developing his idea, and he said he expects Abaca to reach profitability by the middle of next year.
Kirsch insists that Abaca is unlikely to be caught soon.
"Most people like me get 99.8 percent or so with the current volume of users," he said, referring to the percentage of good e-mail he now sees using his system.
"Our performance gets better as we add more users; our competitors already have scale, and we are way ahead even with just 20,000 users. When we get to scale, our performance should be nearly 100 times better than our closest competitor."
In February, Opus One tested six anti-spam products on a stream of 10,000 messages during a 10-day period. Spam-filtering rates ranged from a high of 97.36 percent to a low of 74.10 percent.
"At 99.8 percent you miss two out of 1,000," said Kirsch. "At 95 percent you miss 50 out of 1,000. So other systems give you 25 times as much spam. Who wants that? Nobody we know."
