The researchers estimate that a Wi-Fi attack could take over 20,000 wireless routers in New York City within a two-week period, with most of the infections occurring within the first day.

"The issue is that most of these routers are installed out of the box very insecurely," said Steven Myers, an assistant professor at Indiana University, who published the paper in November, along with researchers from the Institute for Scientific Interchange in Torino, Italy,

The researchers theorize that attack would work by guessing administrative passwords and then instructing the routers to install new worm-like firmware which would in turn cause the infected router to attack other devices in its range.

Because there are so many closely connected Wi-Fi networks in most urban areas, the attack could hop from router to router for many miles in some cities.

The team used what is known as the Susceptible Infected Removed (SIR) model to track the growth of this attack. This methodology is typically used to estimate things like influenza outbreaks, but it has also been used to predict things like computer virus infections, Myers said.

Although the researchers did not develop any attack code that would be used to carry out this infection, they believe it would be possible to write code that guessed default passwords by first entering the default administrative passwords that shipped with the router, and then by trying a list of one million commonly used passwords, one after the other. They believe that 36 percent of passwords can be guessed using this technique.

Even some routers that use encryption could be cracked, if they use the popular WEP (Wired Equivalent Privacy) algorithm, which security experts have been able to crack for years now. Routers that were encrypted using the more-secure WPA (Wi-Fi Protected Access) standard were considered impossible to infect, Myers said.

Myers' model is based on data compiled from the Wireless Geographic Logging Engine (WiGLE), a volunteer-run effort to map Wi-Fi networks around the world, which has over 10 million networks in its database.

Using this data, they were able to map out large networks of made out of Wi-Fi routers that were each no more than 45 meters (49 yards) from the network -- in other words, close enough for an infection to spread. The largest such network in New York included 36,807 systems; in Boston it was 15,899; and in Chicago: 50,084.