Inventor says U.S. security can be monitored online



His idea is to pay spotters $10 an hour to view Web cams from home.
SAN JOSE, Calif. (AP) -- Jay Walker jump-started an online shopping craze by inventing Priceline.com, the Web site that lets people bid on airplane tickets and hotel rooms.
Now Walker is hoping his newest brainchild revolutionizes a completely different field: national security.
The premise behind Walker's USHomeGuard is simple: America has 47,000 power plants, airports and other "critical infrastructure facilities."
Walker believes a terrorist can get within 100 feet of most of them, unchallenged and undetected, and kill or injure thousands.
But if onsite cameras beamed photos to the World Wide Web, Americans could monitor these sites from home. If they spied a potential attacker -- a masked man trying to scale a power plant fence, or a van parked next to a reservoir -- they could alert security agents with a click of the mouse.
Walker envisions spotters getting up to $10 per hour, paid by the government agencies and companies that need protecting. He wants to sell USHomeGuard to the federal government for $1, then charge fees to run the system.
Critics dismiss USHomeGuard as a doomed scheme that exploits Sept. 11 paranoia. Others question the effectiveness of a security system built on the Internet -- itself vulnerable to hackers, power outages and congestion.
David Wray, spokesman for the Department of Homeland Security, said federal officials have not done any "serious evaluation" of the project.
How it works
Walker wants to distribute surveillance across thousands of computers and the people who use them. He says spotters could register online and get paid for clicking through photos.
The spotters answer a simple question about each image: Does it contain a person or vehicle? If yes, local authorities could be notified in as little as 30 seconds.
Walker said it's possible to guard against errors and attempts to foil the system.
For example, as many as one in 10 photos may be traps. If a spotter clicks "no" on a photo of a masked man airbrushed into a reservoir photo, the software suspends him for three minutes -- without pay. He must requalify by clicking correctly through several test photos.
If a spotter clicks "yes" on an unstaged photo, he triggers a first-stage alert. Software automatically routes the same photo to other spotters, and Web cams mounted near the site of the potential attack site beam more photos to more spotters.
When many spotters click "yes," they trigger a second-stage alert. Security supervisors at a data center review photos from all the Web cams and analyze video from the site.
Supervisors who see a suspicious person can speak to him through the Web cam: "Why are you approaching the reservoir?"
If the trespasser is toting a rod and says he's going fishing, the agent might simply ask him to depart. If he doesn't, the security agent may alert local authorities.
It's unclear whether airports, chemical plants and other sites would buy it. Security experts say recognition software can spot potential attacks more economically and with more accuracy than thousands of Americans getting paid $10 per hour.
"Asking people to make a determination of human or not human based on static images is going to be extremely difficult," said Gary M. Lauder of Atherton, Calif.-based Lauder Partners, who heard Walker's business pitch in February. "A computer could probably do a better job."