DARPA Director Tony Tether
BY RICHARD McCORMACK
The director of the Defense Advanced Research Projects Agency (DARPA) testily confronted two important leaders at a House Science Committee hearing, declaring that the computer science community is misinformed about what's happening at his agency.
After two hours of give-and-take between DARPA Director Tony Tether and Wm. Wulf, president of the National Academy of Engineering and Thomson Leighton, chairman of the President's Information Technology Advisory Committee (PITAC) subcommittee on cybersecurity, Tether said: "You know, I don't get any actionable information from these guys. I get a wringing of hands that the money is gone, but not one person has ever clearly defined the specific research that is not being done, the problems that are not being solved or the progress that is impeded" by a perceived reduction in DARPA funding for long-term research in computer science and cybersecurity. "All the people you hear complaining are only complaining because they are not being funded, but they never tell you what it is they want to do other than get the money."
Tether came to a spirited defense of both DARPA and his tenure as its director. He said the conclusions of a recent PITAC study made famous by a front-page story in the New York Times are "not correct."
In a stand-up presentation before the committee using a wireless mic and PowerPoint slides, Tether said he has been pushing DARPA's comptroller -- to the point of his wanting to quit -- to analyze the agency's financial support for university research. The numbers show that DARPA is not reducing its commitment to university research, said Tether.
"When I heard all of the fuss, I thought maybe there was something going on," Tether explained to committee members and about 50 people in the hearing room. "It wasn't like we were throwing out proposals and refusing to fund people. But there was such a noise I figured something must be going on and so I went and gathered some data."
What he found was that DARPA is providing about $525 million a year -- out of a total budget of almost $3 billion -- to universities. "This doesn't sound to me like an organization that is not friendly to universities," Tether proclaimed. Moreover, funding for computer science has been slowly increasing over the years. "This is not a place that is anti-computer science," he added.
"Well, where did the money go?" Tether asked rhetorically. "I'm as curious as all of you, so I went to the Web sites of all the universities." He found that the classification system for scientific research has changed.
"My God, they aren't funding disciplines any more," he told the Science Committee members. "Every Web site at every university you go to is multi-disciplinary something or other -- they are building multi-disciplinary buildings for bio-info. Universities have discovered that multidisciplinary efforts are the wave of the future. We agree. We absolutely agree because that is the approach we've had for 50 years." As a result, measuring funding for specific disciplines like computer science is no longer valid and is probably misleading.
His findings run counter to those of six major computing research organizations. DARPA support for information technology R&D at universities has declined from $214 million in 2001 to $123 million in 2004, a 40 percent drop, according to a "joint statement of the computing research community" submitted to the Science Committee.
"Recently, DARPA has shifted IT funding away from fundamental research at universities in favor of financing classified work and/or more development-related projects," said Leighton, a founder of Akamai Technologies and professor of applied mathematics at MIT, speaking on behalf of PITAC.
National Academy of Engineering president Wulf, who was in charge of the National Science Foundation's Computer and Information Science & Engineering (CISE) directorate from 1988 to 1990 and who received DARPA funding while at Carnegie Mellon University, told the hearing that he is "deeply concerned about what I believe has happened at DARPA -- a drift toward less ambitious and more incremental research."
The Iraq War has "dramatically accelerated this [trend] to focus on reaping the successes of the past," Wulf said as Tether shook his head in disagreement. "Moreover, while there are many DOD organizations that can reap and develop and that collectively have the bulk of DOD's science and technology budget, there was only one old-style DARPA. And it is gone."
The excuse of the Iraq War isn't sufficient to justify this shift away from long-term, revolutionary research, Wulf continued. "One can only wonder at what the world would be like today if the immediacy of the Vietnam War and the Cold War had diverted [D]ARPA from funding crazy ideas like networking, timesharing, VLSI, graphics, RISC architecture, RAID disk systems, parallel computing -- or any other technologies that are essential to today's computer industry and whose results pay off daily to industry, government, the consumer and the military," said Wulf. "At a time of growing global competition, DARPA's dis-investment in university-based, long-term research is, in my view, a risky game for the country."
DARPA's greatest achievements, such as fostering the creation of the Internet, came "back in the days when they invested in people rather than projects," Wulf added.
Tether refuted these charges. He said the agency hasn't changed in 50 years and that if any of its previous directors, with whom he said he speaks regularly, were standing before the committee "they would all agree that what I'm about to give you is something that describes DARPA."
The agency continues to fund "blue-sky" research in many areas including computer science, but "people who were in computer science in the 1990s have not followed" the important trends that DARPA is now pursuing -- notably the development of cognitive processing.
This DARPA thrust of "making a computer learn you as opposed to you having to learn the computer is a big deal," Tether said. "If you really look at it, the only one who's really doing anything in high-performance computing -- that is really building a computer -- is the DARPA program. It's amazing to have these people say that what we're doing with cognitive computing is near term. They can have their opinion that maybe the money shouldn't go into cognitive research and it ought to go someplace else, but to say that cognitive computing is not long term and...will not [have] a major impact is wrong."
Tether also attempted to refute criticism that his agency is not doing enough to ensure the integrity of computer networks and the Internet. He described how DARPA recently built a test-bed of several hundred networked computers at Rome Labs in New York. It let loose a more virulent version of the "Slammer" worm and "we stopped it cold," he said. "We also let it loose without having our new [protection] technique there and it devastated that computer network in seconds. This work is classified and you might ask, 'Why on earth would that have been classified?' Well, first of all, we took a worm that devastated the world and mutated it into a bigger worm. I wouldn't want many people to know how to do that, okay. We also came up with techniques to stop that worm, but we're still not done with those techniques because if people know what that technique was they could design around it."
The DARPA project is developing a permanent fix that can be "put out in code that nobody can reverse engineer and protect the commercial market," he said.
None of this sat well with Wulf and Leighton.
"The problem with the kind of approach that Dr. Tether is pointing out is that at any given point in time, there are about half-a-million bugs in the Windows Operating System," said Wulf, who is on leave from his position as AT&T professor of engineering and applied science in the Department of Computer Science at University of Virginia. "You only need to exploit one to be able to compromise a system. With the Slammer or a mutated Slammer all you have done is simply react to one way of potentially compromising the system out of potentially half a million others. We will not have secure computer systems based on the current models of security. They will not work; they will always have this problem. They are perimeter defense models. They will always have difficulty if there is a way to penetrate the perimeter.
Leighton said the Slammer virus is old -- it was released in early 2003 -- and that during the last half of 2004 there were 7,000 different viruses released into the Internet. "If the research on defending against viruses and worms is classified and not to be released to the general public to defend themselves, then one can question the value of the research," Leighton said. "If you can't use it to defend yourself because you're afraid the bad guys are going to get around it, then it wasn't a very good thing to do. We are basically defenseless today."
Added Wulf: "There is no security in obscurity."
Leighton said that DARPA has "by and large withdrawn the bulk of its funding for university research in cybersecurity" and that it is "vital" to fund university researchers "who are in the best position to develop entirely new approaches to cybersecurity."
Tether also defended the agency's adoption of yearly progress reports to determine if funding for research is warranted. Wulf said that when DARPA was funding his own research in the 1970s and 1980s, the agency entered into long-term agreements and understood that results might not occur for five or 10 years. "The style that DARPA has gone to more recently is very short-term contracts that are requiring demonstrable deliverables every 12 months," Wulf said. "No matter what the amount of money, that builds in a short-term focus."
Tether shot back: "They really are not deliverables," he stated. "They are really long-range programs. We ask our program managers to ask: 'What is the first thing that has to be true in order for the second thing to happen?' There is no sense to spend money on the second thing if the first thing is not done."
DARPA has labeled this process "go, no-go" decisions, and this focus "frightens some people," said Tether. "It's a little bit to me of maybe what's happening in universities. If you go in as a freshman, you spend four years, give them your money and get a degree and never had to show that you had the knowledge to pass physics one. That was a 'go, no go.' That's all we're talking about here is a way to measure progress."
DARPA is really not in the business of providing large sums of money for universities to conduct unfettered research, Tether added. Funding for basic science (6.1) at the agency is only about $150 million a year, $90 million of which goes to universities in the form of grants. The other $450 million going to universities "is really in creating products and when you get into the business of creating products or the demonstration business, it is not unreasonable to ask, 'What is your plan? How are you going to measure yourself to know that you are making progress?' That is what deliverables are all about. They are negotiated between us and the person doing the work. If they don't want to do that, then they shouldn't try to get into the 6.2 [applied research] or 6.3 [advanced technology development] arena, and that is a different arena for normal universities."
Expressing his frustration with Wulf and Leighton, Tether finished with this statement: "I want to know the specific problem they want to work on that if they got the money -- and if the only thing that is stopping them is getting the money -- that if they can do it it will be fantastic. It could be something that will take 20 years. We're used to that."
Science Committee Chairman Sherwood Boehlert (R-N.Y.) concluded: "I think Dr. Wulf and Dr. Leighton are up to the challenge: so you will respond to the challenge offered by Dr. Tether."
Provide us with feedback on this article, suggestions for people to interview, or potential story ideas.
Receive our no-nonsense, content rich e-mail newsletter. (You will NEVER receive spam.)
Manufacturing & Technology News provides its readers with original reporting on ssues that impact YOU -- your company, your industry and your organization. You and your staff can read every article in the latest issue, plus read all the articles written over the past seven years by becoming a regular reader of Manufacturing & Technology News. The cost of a new subscription is $395 per year.
Scan Back Issues | Reports & Analyses | Comments | About Us | How To Order
Copyright © 2005, Publishers and Producers.