Pages: pp. 25-27
Criminals, spies, and terrorists are always looking for better ways to communicate with one another without detection by police and intelligence officers. These officials fear their foes might turn to steganography, a technique that hides a small message within the code representing an image or other standard computer file.
With this in mind, Iowa State University scientists are developing an innovative application, called Artificial Neural Network Technology for Steganalysis (ANNTS), to find such hidden data. The Midwest Forensics Resource Center—which provides research and support services to crime laboratories and investigative scientists throughout the US Midwest—is backing the project.
Most crime labs currently don't perform steganalysis, but the MFRC's team of reviewers found the Iowa State proposal compelling and thus decided to fund it, said Todd Zdorkowski, the agency's associate director. "This is a fruitful line of investigation that has great potential impact in forensic science," he explained.
There is no hard evidence that criminals are using any of the available free or commercial steganography applications yet, noted Michael Morris, special agent supervisor with Iowa's Internet Crimes Against Children Task Force. "But how do you know when you don't have the tools to look for it?" he asked.
Iowa State professor Clifford Bergman and associate professor Jennifer Davidson have developed mathematical techniques to determine whether a file contains a message hidden steganographically.
The most common steganographic method makes small changes to a JPEG or other image via bits of information that represent the hidden message, which could be plain or encrypted text or another image. Each pixel in an image has at least one visually unimportant bit whose value can be altered almost imperceptibly to accommodate a hidden payload.
ANNTS uses artificial neural networks to statistically scan digital files and then examine them for changes in pixel values to detect the likely presence of steganographically hidden data, noted Bergman.
Figure A new system that researchers designed to recognize digital files containing messages hidden by steganography starts by scanning an image or other file. The system extracts various statistical features of the data in the file and uses an artificial neural network to analyze them. If the results exceed established statistical-significance levels, the system flags the file as probably containing a steganographically hidden message.
For example, he explained, in an image with no steganography, the number of pixels with even values, based on their position on a gray or color scale, and the number with odd values will be very different. In an image with steganographically hidden data, the numbers are almost equal. This provides a simple statistical test for steganography.
ANNTS works with multiple types of neural networks designed to find files hidden by different, even sophisticated, steganographic techniques.
The researchers trained their neural networks to identify hidden data by having them analyze large numbers of steganographically altered files. The technology can accurately identify steganographic images between 85 and 100 percent of the time, according to Bergman. He said making ANNTS consistently more reliable will encourage widespread adoption.
The tool cannot decrypt hidden data, which requires different tools.
The Iowa Department of Criminal Investigation plans to begin field-testing the Iowa State application in the near future, said Morris.
Producing content that will run on the many wireless devices available to users today is a major challenge. To help solve this problem, the World Wide Web Consortium is working on a new technology.
The W3C's Device Independent Authoring Language ( www.w3.org/TR/dial) is a markup language used to make single versions of Web sites and applications available to a wide range of cellular phones, PDAs, and other mobile machines. Systems that implement DIAL recognize devices and adapt the content appropriately to take into account issues such as display size and resolution.
By eliminating the need for authors to create and maintain multiple sites and applications for each type of mobile device, DIAL would save considerable time, effort, and money, noted Rhys Lewis, chair of the W3C's Device Independence Working Group and chief scientist for Volantis Systems, a vendor of mobile-content adaptation software.
Figure Authors use the World Wide Web Consortium's Device Independent Authoring Language to create content for Web sites, frequently including data from a database or other sources. The resulting material goes from a Web or application server to an adaptation processor, which tailors the pages for the types of devices that will read them. DIAL thus make Web sites and applications available to a wide range of cellular phones and other mobile machines. Until now, authors frequently have had to create separate Web sites or applications for different mobile platforms.
Vendors could also avoid creating single lowest-common-denominator versions of content that run on multiple devices but offer limited functionality.
"By reducing the costs associated with creating compelling Web sites and applications, DIAL will help to increase the number and variety of appealing sites available to mobile users and help foster increased use of the Web from mobile devices," Lewis said.
The W3C—with the help of device manufacturers, software vendors, service providers, and content developers and owners, including France Telecom Group, Hewlett-Packard, IBM, Nokia, and NTT DoCoMo—has based DIAL on existing and pending standards.
These include extensible HTML (XHTML) version 2.0—a proposed standard that combines features of XML and HTML—and cascading style sheets—which let content authors and, in some cases, viewers define how Web-page elements will appear.
DIAL also works with Media Queries, a proposed W3C standard that provides precise ways to tailor a document's style.
In addition, DIAL works with the W3C's proposed DISelect (Content Selection for Device Independence) standard to let authors and device owners specify and fine-tune content delivery to multiple mobile platforms. DISelect works based on criteria such as device and network capabilities and even user preferences.
Authors could thus specify that parts of a document shouldn't appear on certain types of devices. For example, a cellular phone, which has a small screen, might display only the headline and first few paragraphs of a news article.
Not everyone says DIAL is a good idea under all circumstances. For example, Hakon Wium Lie, chief technical officer for Opera Software, said he doesn't find DIAL's syntax to be intuitive. In some cases, he said, "It looks like a train wreck."
The Device Independence Working Group hopes the W3C will give final approval to DIAL next year, although further testing, as well as approval of XHTML 2.0, could delay this, according to Lewis.
However, he said, numerous companies are taking steps to comply with DIAL even before final approval.
A US researcher has designed a prototype mobile processor that replaces digital and analog circuits in today's chips with programmable analog circuits. This would reduce radio-based wireless devices' cost, size, and power consumption, while increasing their flexibility and enabling them to use bandwidth more efficiently.
The project by Benjamin Vigoda—a research scientist at Mitsubishi Electric Research Laboratories and research affiliate in the Massachusetts Institute of Technology Media Laboratory—would address the challenges that cell phones, remote sensors, and other mobile devices face in accommodating increasingly power-hungry applications and using wireless networks' limited bandwidth wisely.
Vigoda says his project targeted mobile devices' two most power-intensive and expensive processors: the analog front-end chip that receives the incoming signal and sends it on for further processing, and the digital baseband-signal processor.
The new circuits are reprogrammable, noted Vigoda. He said his chip design is modular and thus gates can be rearranged as necessary for different purposes. This would let manufacturers reuse chips for multiple applications.
The new approach is power efficient, an important consideration for mobile devices, which must operate within batteries' limited resources, Vigoda says.
The programmable-analog circuits could cut power consumption by 10 to 100 times by eliminating the need for existing mobile chips to convert between analog and digital. Moreover, analog devices generally operate more efficiently than their digital counterparts.
The circuits also use network bandwidth more efficiently because they offer better error correction; more complex signal detection and demodulation; and multiple-in, multiple-out (MIMO) technology, which uses multiplexing and two or more transmitting and receiving antennas to increase the available bandwidth.
The prototype chip uses CMOS-based transistors, so companies could make the circuits via standard semiconductor-manufacturing processes.
The flexible processors would yield radios—in some cases called cognitive radios (see Computer, May 2006, p. 22)—whose signals could be changed on the fly and adapted to a particular platform, frequency range, interference problem, or environment.
In addition, their efficiency would enable more complex signals and higher communication rates, which could speed transmissions and downloads.
Vigoda has completed testing of the analog logic gates on his chip. He said two manufacturers have expressed interest in the processor, and he predicted communications systems, including cellular phones, could be using it in five years.
Analyst Will Strauss with Forward Concepts, a market research firm, said it sounds as if the new chip design has potential. However, he noted, cellular technology might change considerably in five years, so it's hard to predict whether the approach will be successful.
Carnegie Mellon University researchers have designed an online game called Phetch with the ultimate goal of making Web sites more accessible to the visually impaired or blind.
CMU assistant professor Luis von Ahn invented Phetch while a postdoctoral student at the school. The multiplayer game—which can be found at www.peekaboom.org/phetch and which operates from a CMU server—works with four contestants at a time.
One participant writes a caption that describes an online image. Each of the others then goes through all related images in a database to look for the one they think best matches the caption. The first player to find the correct image, if any are able to do so, wins.
Multiple games are played for each image to select the best single caption, determined by factors such as how quickly players could find the correct picture based on the text provided.
The CMU system saves the best descriptions, along with the corresponding images, in plain text so that a visually impaired person's system could convert data easily into Java, XML, or whichever format it uses.
Before the research team commercializes the technology, it must develop a tool bar or other technology that can connect visually impaired users to the CMU server so that they can access images and captions via screen readers.
In one week of testing, 130 Phetch players generated 1,400 captions. He said the research team hopes to start by producing captions for at least 1 million images.