Dr. Dobb's Journal August 2002
Biometric security devices, often touted as a next-generation solution for systems requiring high security, suffered a credibility blow when a Japanese cryptographer proved that many systems relying on fingerprint recognition can be fooled by cheap, easily manufactured fakes. Tsutomu Matsumoto was able to lift latent fingerprints left on glass, digitally photograph them, and make a three-dimensional model using a photosensitive printed-circuit board. He then used the mold to form a fake finger out of common household gelatin. The gummy finger was tested against 11 fingerprint recognition systems currently sold; it passed about 80 percent of the time. Matsumoto was also able to defeat some "living finger detection" measures simply by moistening the fake finger.
"The results are enough to scrap the systems completely, and to send the various fingerprint biometric companies packing," according to Bruce Schneier in his Crypto-Gram newsletter (http://www.counterpane .com/crypto-gram.html).
Matsumoto presented his findings at the Optical Security and Counterfeit Deterrence Techniques conference; copies are available upon request from tsutomu@mlab.jks.ynu.ac.jp.
A Beowulf cluster running on Transmeta Crusoe blades, unveiled at Los Alamos Labs, sits in a hot warehouse with no special cooling equipment or ventilation, and it almost never crashes. The 240-node system, called "Green Destiny," was designed by the Research and Development in Advanced Network Technology (RADIANT) group, headed by Wu-chun Feng, and is intended to show the power of "supercomputing in small spaces."
In his trials with Green Destiny's precursor, a 24-node cluster, Feng has found that "Bladed Beowulf" systems give about 75 percent performance when compared to an Intel system. The cost savings, however, are huge. When Feng estimated the price of housing and maintaining two supercomputers built on the different architectures, he found that the Bladed Beowulf was three times cheaper. In the final price/performance analysis, the bladed system is, therefore, twice as attractive.
"The continued tracking of Moore's Law will result in the microprocessor of 2010 having over one billion transistors and dissipating over one kilowatt of thermal energy; this is considerably more energy per square centimeter than even a nuclear reactor," wrote Feng (http://public.lanl.gov/feng/Bladed-Beowulf.pdf). "The Bladed Beowulf simply cannot compete with ASCI-style supercomputers due to their massive compute and communication capabilities. However, this research can perhaps be viewed as the foundation for the supercomputer of 2010."
SETI@Home (http://setiathome.ssl.berkeley .edu/), the distributed computing project that seeks to find evidence of extra-terrestrial communications, returned its 500 millionth result recently. The project is the one of the most popular, longest running distributed computing efforts: It has attracted 3.5 million users and celebrated a three-year anniversary in May.
The SETI@Home project has a receiver that piggybacks on the Arecibo antenna in Puerto Rico, listening for about a second each to random spots in the sky. Enormous amounts of data are collected this way, filling about one 35-gigabyte DLT tape every day. The tapes then have to be flown to Berkeley, where the data is broken up into work units and uploaded to the SETI@Home servers. Early on, the project found that some users were falsifying data in order to return interesting signals; now each unit is sent out to multiple users, so the results can be checked against each other.
SETI@Home's greatest challenge may be its own popularity. The project, which consumes about 40 percent of the University of Berkeley's total bandwidth, was choked off for a few days in January when the university's own bandwidth consumption suddenly fluctuated upward. SETI@Home is now struggling to create an independent pipeline sufficient for all its users.
The Buzz Robotic team from Connecticut's Enrico Fermi High School took the grand prize in this year's FIRST Robotics National Competition (http://www .usfirst.org/), a challenge that gives high-school teams six weeks to design and build a robot using a standard kit of parts. The competition's top prize comes with a $10,000 scholarship for one member of the winning team.