Researchers question Carnivore review

'Show us the source code'

A blue-ribbon panel of computer scientists and researchers added fuel to the controversy over the FBI's Carnivore on-line surveillance tool on Monday, calling a recent review of the system inadequate.

While crediting the review team with a "good-faith effort" at independent analysis, and lauding the Justice Department's openness in handling the review, the five scientists concluded in written comments that "the limited nature of the analysis described in the draft report simply cannot support a conclusion that Carnivore is correct, safe, or always consistent with legal limitations."

The comments were by researchers Steven Bellovin and Matt Blaze from AT&T Laboratories, University of Pennsylvania computer science professor David Farber, SRI International principal scientist Peter Neumann and computer security expert Eugene Spafford from Purdue University.

"Although the report is useful as far as it goes, there are a lot of issues that remain to be discussed," said Neumann in an interview. "In particular the operational issues and the questions regarding auditing are important."

Among other recommendations, the researchers suggested the Department of Justice release Carnivore's source code for public review.

At issue is a report the Justice Department commissioned from the IIT Research Institute (IITRI) in September after advocacy groups and lawmakers voiced concerns that Carnivore, a tool designed to facilitate court-authorized Internet surveillance, might violate the privacy of innocent Internet users or interfere with the operation of ISPs.

Late last month the Justice Department released a draft of the 121-page report for a public comment period, which ended Friday.

The five scientists became involved in the Carnivore controversy on 2 October, when Justice officials summoned them to Washington for a meeting on the pending IITRI review. "It was part of our community outreach to talk to people," said Justice Department chief science and technology officer Donald Prosnitz. "They are very prominent people, well known in the industry... It was sort of a brainstorming session to get their suggestions."

According to Prosnitz and an IITRI researcher, some of those suggestions for putting Carnivore under a microscope were reflected in the draft report. But the five scientists charge that the analysis didn't examine Carnivore's code for common bugs like buffer overflows; gave little attention to the interaction between Carnivore's software and the Windows operating systems on which it runs; didn't adequately discuss the system's audit trails; and did not test Carnivore's ability to handle RADIUS, the protocol that ties a user's identity to an Internet address.

Those who are concerned that the system produces correct evidence, represents no threat to the networks on which it is installed, or complies with the scope of court orders should not take much comfort from the analysis described in the report or its conclusions," the scientists conclude.

"That's pretty strong language that they use," said Henry Perritt, dean of the Illinois Institute of Technology's Chicago-Kent College of Law, and a member of the review team. "It's almost like saying we can't conclude that Carnivore is perfect, and I certainly agree with that."

Perritt called the comments "extremely helpful," but defended IITRI's report, which was produced under a tight deadline. "We've acknowledged in the draft report specific places where time limitations only permitted us to do less than ideal testing," said Perritt. "In the limited time available to turn the draft report into a final report, we're going to respond to as many of those concerns as we can."

The revised version of the Justice-Department-approved Carnivore report is due Friday.

© 2000 SecurityFocus.com. All rights reserved.

Related Stories

Congress peers deeper into Carnivore
FBI's Carnivore review is mixed

Sponsored: Driving business with continuous operational intelligence