To reduce the heat that chips generate—and thereby improve performance and decrease power consumption—industry researchers for the first time have integrated onto a processor a small thermoelectric device that cools only areas where hot spots occur.

The scientists—from Intel; research institute RTI International; and RTI spin-off Nextreme Thermal Solutions, which manufactures microscale thermal and power-management products—developed Embedded Thermoelectric Cooler technology.

The eTEC approach uses electrical current to cool targeted parts of a processor by about 15 degrees Celsius.

When processors operate, they generate heat. Excessive amounts can hurt performance and reliability, and even damage circuitry, noted Tom R. Halfhill, senior analyst with market research firm In-Stat's Microprocessor Report.

Current cooling techniques, such as fans used with heat sinks, take up too much valuable real estate on chips, which is a particular problem for processors used in small devices, noted Paul A. Magill, Nextreme's vice president for marketing and business development. The approaches are also inefficient because they pull heat from the entire chip rather than just the hot spots, he explained.

Rather than remove heat uniformly across a chip, the new technique removes heat only at hot spots. According to Magill, research indicates that processor speed is limited not by a chip's overall heat level but by the levels at the hottest spots.

The researchers used thermoelectric coolers made from tiny films of bismuth telluride between 5 and 15 microns thick that attach to, rather than sit on, the processor. They thus occupy only 10 percent of the chip area.

The temperature difference between the cooling device and the warm chip generates an electric current. When the current flows through the two different materials, the cooler element, in essence, removes heat from the chip and passes it to a heat sink.

The device offers 10 to 20 times the cooling capability of conventional systems because bismuth telluride absorbs heat effectively, said Magill. This process is more efficient and eliminates the need for large, on-chip devices such as fans or pumps, which consume considerable energy, Magill explained.

Multiple coolers are used at pre-identified hot spots on the same chip. Microprocessor and microcontroller designers know the locations of hot spots, which are typically in the areas where calculations take place toward the middle of the chip.

Manufacturers don't necessarily know where hot spots will be on field-programmable gate arrays, as users can configure FPGAs in different ways.

A concern for the researchers is that the material used to attach the cooler to the chip surface could disrupt heat removal. They are thus exploring connectors such as special types of solder and carbon nanotubes, and will choose the approach that works best with their manufacturing process, according to Magill.

The Internet Corporation for Assigned Names and Numbers has begun letting domain-name registries in countries and territories that use non-Latin alphabets apply for permission to have full domain names appear in their native languages.

ICANN, responsible for managing the assignment of domain names and IP addresses, says this will increase Internet accessibility worldwide, particularly to people using non-Latin alphabets such as Arabic, Chinese, Cyrillic, Greek, Hebrew, Japanese, and Korean. Most Internet users work with non-Latin alphabets, noted Tina Dam, ICANN's senior director for internationalized domain names.

The agency, which has tested its IDN plan for several years, calls this the biggest technical change to how the Internet works since its invention.

When the domain name system (DNS) was introduced in 1983, it worked with only 37 of ASCII's 128 characters: the Latin alphabet letters A through Z, the numbers 0 through 9, and a hyphen, noted Dam. Unicode—which currently has about 107,000 characters, including those from non-Latin alphabets—did not exist then.

In the past, Internet officials discussed making domain names available in multiple alphabets, Dam said, but there was no practical solution for this. However, the issue has resurfaced with growing global Internet use.

With ICANN's system, an application—such as a browser or e-mail system—that wants to access the Internet via an IDN converts non-ASCII Unicode characters to ASCII characters before performing a DNS lookup, explained Ben Wilder, senior software engineer for domain-name registry 101Domain Inc.

In addition, the DNS itself—which translates domain names written with characters and digits into Internet Protocol numerical addresses—will store IDNs as sequences of ASCII characters beginning with "xn--." This will maintain compatibility with the current ASCII-based DNS.

A domain name such as the Swedish "räksmörgås" would become "xn--rksmrgs-5wao1o," Dam explained.

The plan's technical challenge will be changing application software to support IDNs, noted Wilder. Applications would support IDNs either natively or via plug-ins, he explained. He said Mozilla's Firefox and Microsoft's Internet Explorer, Word, and Outlook already will work with IDNs.

Initially, ICANN will allow full IDNs in one non-Latin alphabet per country. The alphabet must be an official language of a country or territory, used for formal communications.

ICANN has begun a fast-track application process and has received 16 requests for IDNs, which could become available this year, Dam said.

Some security experts warn that IDNs could make fighting cyberattacks, including malicious redirects of website visitors, more difficult.

Also, companies seeking protection of their brand names in different languages will probably have to register many more domain names, noted Susan Jessup, 101Domain's director and registry liaison.

News Briefs written by Linda Dailey Paulson, a freelance technology writer based in Portland, Oregon. Contact her at ldpaulson@yahoo.com.

An academic researcher has developed a technique that automatically generates filters to protect applications against malicious exploits until software vendors develop and release patches, thereby reducing critical delays in defending systems.

Traditionally, antivirus vendors don't develop a filter for an exploit until it appears and starts causing problems. They thus react to each attack, virus, worm, or Trojan one by one and only after a delay. This could let attacks spread quickly and present a particular problem for sensitive applications in security-critical industries such as financial services.

University of California, Berkeley, associate professor Dawn Song's BitBlaze project, including the BitBlaze Binary Analysis Platform, is designed to address these issues.

In her multifunctional platform, explained Song, one application combines dynamic and static analysis with symbolic reasoning and program verification to examine binaries and identify vulnerabilities and ways they could be exploited. Another application detects new attacks on vulnerable programs.

When the system detects a new attack, it performs an analysis to rapidly determine the vulnerability that permitted the incident and immediately and automatically generates filters. The filters protect against the current exploit—as well as future attacks that try to take advantage of the same vulnerability—until application vendors develop and release patches.

Song has worked on her research with several technology companies, including Google and Symantec. However, representatives from these companies said they haven't incorporated this specific work into products at this time.

The ability of the UC Berkeley technique to evolve defenses dynamically is new and interesting, said Graham Titterington, principal analyst with Ovum, a market research firm.

According to Chenxi Wang, principal analyst with Forrester Research, the new approach might have limited applicability because malware increasingly relies on social engineering—enticing users to download a harmful file—rather than on software vulnerabilities.

Song said her research team has released BitBlaze as open source software, available at http://bitblaze.cs.berkeley.edu/release/index.html.

The platform could function either as a stand-alone application or be incorporated into existing applications such as malware-analysis engines or intrusion-detection and -prevention systems.

Song is working on a new project called WebBlaze ( http://webblaze.cs.berkeley.edu), designed to develop new technologies for securing the Web via BitBlaze-like approaches. Some browsers, such as Google's Chrome, have already incorporated some WebBlaze technology, she noted.

The UC Berkeley research team is planning to build a production-quality version of its technology that could be commercialized.

Wang said the technique would be useful primarily in the consumer market because businesses probably wouldn't want to pay extra for a security application that generates filters for as-yet-unseen threats. She also expressed uncertainty about what business model could work for vendors.