Home' Technology Review : April 2005 Contents 83
EDITED BY MONYA BAKER
Each month brings new investigative tools, new ideas for revolutionary
technology, and revolutionary applications of existing technology.
No one can know today which will matter most tomorrow.
But these represent Technology Review s best prediction.
Hackers strike mobile phones
: Computer viruses and worms
can be sent over mobile wireless networks
almost as easily as text and voice messages,
and any device that receives voice or data
digitally is vulnerable. Some programs
drain devices batteries, disable buttons,
or assail users with mobile spam; the more
malicious ones steal infor mation. David
Dagon and his colleagues at the Georgia
Institute of Technology and Virginia Poly-
technic Institute and State University have
created a taxonomy (a systematic classi -
cation) of mobile "malware" threats.
: By sorting mal-
ware according to how it works, the tax-
onomy shows not just what kinds of attacks
have occurred but also what kinds are pos-
sible. Its categories include the v ulnerabili-
ties that malware exploits (say, certain
layers in a routing network) and the types
of problems it causes. All existing malware
causes semantic errors, a type of error that
orders the mobile system to misbehave.
The taxonomy shows that new attacks
might exploit another class of errors, syn-
tax errors, that confuse the phone by issu-
ing orders it can t understand, causing the
cell-phone equivalent of Microsoft s Blue
Screen of Death.
: While mobile antivirus
strategies will draw from their desktop
counterparts, mobile protection algo-
rithms will need to be optimized for the
lower CPU usage, higher power e ciency,
and other idiosyncrasies of small devices.
To prepare the best defense, engineers
and end users need a map of the routes
the enemy might take. That is what the
Source: Dagon, D., et al. 2004. Mobile phones as
computing devices: the viruses are coming! IEEE
Per vasive Computing 3:11--15.
Vision algorithm models
: To locate an object in an image,
computer-vision algorithms often use
mathematical models of the object s shape.
But nding the boundaries of "deform-
able" objects, like human organs, is di -
cult, because the model must account for
all of the objects potential shape altera-
tions. Algorithms that can pick out the
edges of stretched or squashed objects are
often ine cient; they require users or
other algorithms to provide initial esti-
mates of the objects positions and orienta-
tions. Pedro Felzenszwalb of the University
of Chicago has developed a deformable-
shape model that helps locate such objects
in images quickly and accurately.
: Felzenszwalb s
algorithm is able to represent any two-
dimensional shape that contains no holes.
Each shape is modeled by a collection of
triangles that approximates the boundary
of the undeformed shape. The algorithm
assumes that some triangles can be dis-
torted more than others, and that triangle
edges at the boundaries of an object tend
to coincide with changes in image bright-
ness. To match the model to objects in the
image, the algorithm deletes one triangle
at a time from the model, transferring the
information about its best- tting defor-
mations and image locations to a neigh-
boring triangle. Once all the triangles are
eliminated, the stored information can be
used to quickly decide the area in the im-
age that best matches the model. Thus,
the algorithm can nd the object without
searching for every possible location, ori-
entation, or deformation of the model.
Given information about how an ob-
ject could be represented by triangles,
the algorithm nds the object s boundar-
ies in the image. Given a set of example
shapes, the algorithm can also constr uct
a general model for a class of objects, such
as hands or leaves.
: Better modeling of
deformable shapes increases the range of
objects that computers running vision al-
gorithms are able to automatically recog-
nize. Felzenszwalb s method could thus
be important for applications such as
medical imaging and surveillance. It is as
accurate as the leading methods for nd-
ing object boundaries in medical images,
Links Archive March 2005 May 2005 Navigation Previous Page Next Page