A committee of scholars agreed yesterday to devise a system to limit the publication of computer research on top-secret codes that the government says could endanger national security.

The action was believed to constitute the first time that an academic group has advanced a proposal that could result in prior restraint of free and open publication of research.

The committee, the Public Cryptography Study Group, was formed by the American Council on Education at the request of the Defense Department's National Security Agency to develop the mechanism in order to alleviate the agency's growing fears over cryptographic research.

It considered the problem for a year and rejected the government's original assertion that legislation was needed to block publication of cryptographic research. Instead, the committee agreed to propose an experimental voluntary plan, under which computer scientists would submit their work for review by the NSA before publishing it.

Although no formal vote was taken, the committee's cochairman, Ira Michael Heyman, chancellor of the University of California at Berkeley, canvassed the nine members informally and declared, "We've got what appears to be a consensus."

The only dissent came from Prof. George I. Davida of the Georgia Institute of Technology, who said in a minority report: "I find NSA's effort to control cryptography to be unnecessary, divisive, wasteful and chilling."

The government says that national security is at stake because research in this area has progressed to the point that full knowledge of it could help other countries in breaking U.S. codes and ciphers, and could alert unfriendly nations that their ciphers are insecure.

The report now goes to the American Council on Education, which includes 1,300 colleges and universities and 150 other academic organizations. Under the voluntary proposal, it will be up to each researcher and each scholarly journal to decide whether to comply.

Under the plan, the final wording of which is to be worked out today, a research paper in cryptography would be forwarded to the security agency before publication. The agency could approve it as written, or ask for changes. The author would be free to accept the changes or go ahead with publication.

In the event of a disagreement, the paper would be sent to an advisory committee. Of its five members, two would be appointed by the NSA director and three by the president's science adviser from a list provided by the head of the National Academy of Sciences.

The advisory committee, whose members would carry the highest security clearance, would be told exactly what the security agency's objections were. They could then agree either with the author or with the agency.

In any event, the author and the scholarly journal would be free to publish the original paper or to accede to the changes at their own discretion. p