SAN FRANCISCO (AP) _ A survey that tried to measure the extent of computer break-ins came up with insufficient data, apparently because companies are reluctant to disclose whether they have been victimized.

Of 4,971 questionnaires sent out, only 428 responses were received. Of the respondents, more than a third reported computer system break-ins within the past year, and half the intrusions were inside jobs.

The survey being released today was sent to members of the Computer Security Institution, a San Francisco-based association of information security professionals.

The low response rate means results are not statistically meaningful, said Ann Kalinowski, a statistician with Failure Analysis Associates in Menlo Park, Calif.

The figures would not reflect intrusions that are not detected or not reported _ or the possibility that those who did not respond to the survey have higher or lower rates of computer break-ins.

Few respondents said they would report to law enforcement agencies if they thought they had been victimized _ most cited fear of negative publicity.

``If that's accurate, it causes considerable concern,'' said FBI spokesman George Grotz. That tells us we've got to do a better job of educating the public as to our responsibilities. It's a wake-up call.''

The FBI's International Computer Crime Squad's San Francisco office helped write the survey, which Grotz defended as a ``first step,'' despite its imprecision.

``A lot of computer crime and fraud goes undetected, or unreported,'' said Fred J. Rica, of Price Waterhouse LLP's information system risk management group in Morristown, N.J. Companies are very hesitant to publish figures saying, `Yeah, we've been hit for X number of dollars.'''

A separate survey also being released for publication today made another attempt to estimate the number of U.S. adults who use the Internet.

The Harris Poll said its nationwide survey in late April indicated 33 million adults use an online service, 41 million use electronic mail and 29 million use the Internet. Some researchers believe such polls measure too liberally by letting the respondent define what using a computer entails.