key to Net viruses
Technology Research News
When the Internet linked distant computers
30 years ago, its founders were probably not thinking about protecting
the machines from infecting each other. Today's exponentially larger Internet,
however, is vulnerable to software viruses in much the same way that large,
crowded human populations are more likely to fall prey to biological viruses.
The Internet has
a scale-free structure, meaning it has a few pages, or nodes with many
connections to other pages and many with just a few connections. Researchers
looking into how bits of disruptive code spread on the Internet have found
that this structure isn't conducive to the conventional practices of inoculating
The researchers did, however, find an inoculation strategy that promises
to protect computers more effectively.
When they applied an immunization strategy that's commonly used for biological
populations to a simulated scale-free network, it simply didn't work,
said Alessandro Vespignani, a research scientist at the Abdus Salaam International
Center for Theoretical Physics in Italy.
The researchers inoculated progressively larger numbers of nodes, expecting
the epidemic to eventually die out, he said. It did not even when they
inoculated more than 90 percent of the nodes, he said. "Surprisingly,
in scale-free networks we observed that infection survived... in the presence
of massive vaccination campaigns involving the majority of the population.
We realized that random... schemes were practically useless in scale-free
The Internet is generally more vulnerable than human populations because
the connections among computers are both more numerous and structured
differently than many of the human connections that allow viruses to spread.
Scale-free networks have some nodes -- large portals, for instance --
that contain more connections to other pages than even the most widely-traveled
people could possibly have with other people.
The researchers eventually caused the epidemic to die out by targeting
nodes that had a high number of connections rather than inoculating individuals
Using this scheme, the researchers sharply lowered the network's vulnerability
to epidemic attacks, Vespignani said. "We have tested this recipe on a
real map of the Internet [with] a targeted immunization involving all
the most-connected individuals. In this case, by immunizing [less] than
one percent of the total population, the cyber infection cannot propagate,"
The research explains why, though antivirus software is very successful
in protecting individual computers, it does not prevent computer infection
from becoming endemic. "The 'I love you' virus is still in the top list
of most frequent viruses more than a year after its introduction... because
the global implementation of antivirus [software] is practically equivalent
to a random... vaccination," Vespignani said.
Ironically, this scheme could also be useful in the biological world where
some of the paths viruses take to propagate in a human population have
some similarities to the Internet. A map of human sexual relations, for
instance, has scale-free properties, said Vespignani. The research implies
that epidemics spread this way could be prevented more effectively by
targeted vaccination of the few promiscuous individuals, he said.
This type of targeted vaccination would also prove to be much cheaper
than the random kind, Vespignani said. "Instead of massive vaccination
campaigns, we can think of identifying the network connectivity hierarchy."
Controlling the hubs that spread the infection more quickly is both more
effective and requires relatively few inoculations, he said. "The strategy
is... particularly convenient in terms of economical and practical resources."
The problem in both the Internet and biological networks that harbor a
scale-free nature is identifying the large hubs, said Vespignani. "The
difficulty... is... detailed knowledge of the network connectivity. This
is not always possible for privacy and economical reasons. It is very
difficult to obtain a complete map of the Internet because many providers
do not want to share publicly their information. As well in the case of
sexual diseases we have to rely on people's concerns about their own sexual
habits," he said.
This strategy "looks reasonable. It is consistent with my experience,"
said Gene Spafford, a computer science professor at Purdue University.
"I'm surprised no one else has noted this property in research... either
in networks or in epidemiology," he said.
One complication that the model leaves out is the notion of workgroups,
or local area networks where each machine is connected to all the other
machines in that group, and an infection of one infects all the others,
It is hard to estimate when the research could be used to actually inoculate
networks, said Vespignani. "The use of these results is strictly related
to social factors -- individuals' privacy -- and the existence of control
agencies." These make estimating the time frame difficult, he said.
Vespignani's research colleague was Romualdo Pastor-Satorras of the Technical
University of Catalonia in Spain. The research was funded by the European
Community, the Spanish Ministry of Education and Culture, the Abdus Salaam
International Center for Theoretical Physics (ICTP) and the Technical
University of Catalonia (UPC).
Funding: Government, Private
TRN Categories: Internet
Story Type: News
Related Elements: Technical paper, "Optimal Immunization
of Complex Networks," posted in the Los Alamos physics archive at arXiv.org/abs/cond-mat/0107066
Hubs key to Net viruses
water spins gold into wire
Virtual reality gets easier
Laser emits linked photons
Dye brightens micromachines
Research News Roundup
Research Watch blog
View from the High Ground Q&A
How It Works
News | Blog
Buy an ad link