From The Lab: Information Technology

Kill the Bots!
Software thwarts malicious hackers

Context: The malicious computer programs known as “worms” infect more than 30,000 new computers every day. Unbeknownst to their owners, the compromised machines follow orders to send spam, say, or to access particular websites. If enough of these so-called zombie machines simultaneously contact a particular Web server, they can knock it out of commission. Professional hackers have used the threat of such “distributed denial-of-service attacks” to extort money from businesses. Last year, one company’s business manager was indicted for paying hackers to use zombies to take down competitors’ websites. The zombies dodge a Web server’s defenses by disguising themselves as legitimate users and then block access to the server by overloading not only its network bandwidth, but also its CPU, memory, disk space, and database resources. Now, led by Dina Katabi, researchers from MIT, Princeton University, and Akamai Technologies have developed Kill-Bots, a clever, simple, and cheap means of distinguishing friend from foe. Unlike other products, it allocates a server’s system resources only after a user is confirmed as legitimate.

Methods and Results: Kill-Bots, a software modification to a server’s operating system, kicks in whenever a website is in danger of being overwhelmed by traffic. The software asks requesters to solve a simple graphical puzzle before it grants access to server resources like buffer space. Humans can solve these puzzles easily; zombies cannot do so at all. Addresses that repeatedly request site access without solving the puzzle are blacklisted automatically. When the load on the Web server decreases, it stops issuing puzzles and accepts requests from nonblacklisted addresses, so even real users who did not solve the puzzle can gain access.
In experiments, a Kill-Bots-protected Web server successfully endured five times as many hits as an unprotected Web server. Not only did the Web server stay online, but protected websites also maintained speedy response times, even during the height of the attack.

Why it Matters: Worries over dis­tributed denial-of-service attacks are spreading. Most Web server defenses use authentication procedures that are easily outwitted and depend on replicated content, mul­tiple CPUs, and extra bandwidth, all of which cost money. Kill-Bots is much cheaper and can be easily deployed; it requires no changes in users’ Web browsers and works with the very large number of Web servers running Linux. Although Kill-Bots occasionally misclassifies legitimate users as zombies, it allows websites under attack to remain available and so promises to keep the Web open for business, while barring the way for thieves and vandals.

Source: Kandula, S., et al. 2005. Botz-4-Sale: surviving organized DDoS attacks that mimic flash crowds. Paper presented at 2nd Symposium on Networked Systems Design and Implementation. May 2–4. Boston, MA.

Dethroning the Transistor
A new molecular logic switch

Context: The terms “semiconductor” and “computer” have become entwined; better semiconductor manufacturing has enabled the release of chips with smaller and faster circuits every year. But in a decade, the miniaturization of silicon transistors may reach physical limits that prevent further improvements. So engineers from Hewlett-Packard have created a molecular device that could be the heart of the computer of the future.

Methods and Results: The circuits proposed by Phil Kuekes and his HP colleagues rely on a “crossbar”: an array of crossed metal wires separated by a single layer of molecules. Like a transistor, a crossbar can be switched between a high and low conducting state, allowing it to store information. Kuekes shows how to link crossbars so that they can not only store data but also restore noisy data and apply a logic operation called inversion, which swaps binary 0s for 1s and 1s for 0s. The crossbars can be linked with other components to generate the entire family of logic needed for computing. The researchers have yet to combine all these capabilities into a stand-alone computing device, and they have not yet found a way to make molecular junctions that switch states quickly and reliably enough to compete with silicon transistors. Nonetheless, they have provided the first demonstration that crossbars can perform all the functions transistors can perform.

Why it Matters: The HP researchers have cleared a path toward a computer chip without conventional transistors. The process used to create their crossbars is inexpensive and in principle could lead to logic elements even smaller than those constructed from the most advanced silicon transistors, which would enable faster and more efficient computer chips. But even if the performance and reliability of crossbars surpass those of transistors, they may still lack the muscle to compete with the entrenched semiconductor industry. Crossbars may instead find their first applications elsewhere, in flexible logic devices, for example, or displays.

Source: Kuekes, P. J., D. R. Stewart, and R. S. Williams. 2005. The crossbar latch: logic value storage, restoration, and inversion in crossbar circuits. Journal of Applied Physics 97:034301.

Brighter Silicon
Toward more-efficient optical devices

Context: Silicon is good at shuttling electrons around chips but much worse than most other semiconductors at manipulating light. This shortcoming has kept optical chips, which transmit information more efficiently than electrical chips, from wider use. Silicon “nanocrystals,” a few atoms of silicon covered with an oxide layer, emit light more efficiently than bulk silicon, but devices incorporating them wear out quickly and are still too inefficient for most applications. Now, a team led by Harry Atwater of Caltech has improved silicon’s ability to emit light, giving a boost to an industry looking for new ways to make faster chips.

Methods and Results: In a conventional light-emitting diode (LED), electrons traveling through a semiconducting crystal meet electron “holes” – or gaps left in the crystal by absent electrons – and lose energy, which is emitted as light. But this approach doesn’t work well with silicon nanocrystal LEDs, where electrons moving toward the holes can collide with atoms in the crystal and displace them, degrading performance.
Previous silicon LEDs used separate electrodes to inject holes and electrons into silicon nanocrystals. But Atwater and colleagues figured out how to inject both from a single electrode. In their device, a thin layer of silicon nanocrystals sits atop an electrode that alternates between adding electrons and adding holes. This keeps electrons from rocketing violently across the crystal and damaging it. Also, by elimi­nating one of the entry and exit points for electrons, the Caltech group has made devices that are easier to fabricate and more consistent in performance.

Why It Matters: The new LED can be built using standard equipment that could be integrated into a chip-manufacturing line. Its performance, however, is still low enough to limit its use. To improve the processing speed of silicon chips, the LED would have to switch on and off more quickly; to be of use in a display, it would have to consume less power. Nonetheless, the semiconductor industry has much practice improving the performance of silicon chips. The problems of speed and power may not remain unsolved for long.

Source: Walters, R. J., G. I. Bourianoff, and H. A. Atwater. 2005. Field-effect electroluminescence in silicon nanocrystals. Nature Materials 4:143–146.