In 2006, when I discovered that my request to review the CT/AT issue had been killed as soon as I turned my back, I asked George to reexamine the issue independently. It was simple request, but it chafed some of the people who had done the original testing at the then-FAA labs. They may have seen it as a political infringement on pure science, but to me, it looked like the lab’s insulated existence had an unintended dark side. The scientists were judge and jury for all TSA security technology. There was no external peer review; infallibility was assumed. And not only was their network a closed circuit, they didn’t incorporate the agency’s intelligence resources into their work to understand the true nature of the threat. So, when a decision had to be made about how to replace the machines that screened carry-on baggage at checkpoints, I was a bit wary of the lab’s insistence that CT was the only way to go.
The checkpoint x-rays that the TSA inherited had first been deployed in the 1960s. Known as TRXs, they relied on a single x-ray source on the top of the box and produced grainy images. We desperately needed to upgrade them, and a kind of AT technology, based on multiple-view x-rays and known as ATXs, was available. Aside from their complete imaging of bag contents, I liked the flexibility of the ATXs because they were capable of accepting new search algorithms. Any time the terrorists tweaked explosives formulas, we could just swap out the scanning software instead of buying a whole new machine. The FAA lab, however, was pushing for the more sophisticated CT-based solution at our already overcrowded checkpoints. These machines had many of the same problems associated with their larger cousins that scanned checked baggage. They cost much more and could process only about half the 400 bags an hour necessary at our checkpoints. Moreover, their bulk completely blocked the TSOs’ view of the passengers they were screening. And they were not ready to be installed.
Seeing a steady stream of aviation plotting, I didn’t want to wait for the CTs to be perfected. By this point, I also wanted a second opinion, so I asked George to go back to the larger scientific community and his impressive range of contacts. George called up a guy he knew at the Department of Defense, who suggested doing our testing at an old hangar at Tyndal Air Force Base, just east of Panama City, Florida. Three weeks later, George and some scientists from the Lawrence Livermore national lab and DHS Science & Technology office were set up at a site on the Panhandle, isolated enough that if something went wrong, it wouldn’t draw attention. Besides, the Air Force tested bombs there.
A few days later, as they were running different hydrogen-peroxide concoctions through the AT machines, testing for possible differences in their electronic signatures, a researcher yelled out, “Wait! It looks different from water!” The hydrogen peroxide in that machine did look slightly different from water. Eventually the AT was calibrated to differentiate hydrogen peroxide from wine, water, shaving cream, and contact-lens solution—and, as margin for error or if terrorists tried using many small bottles, to detect it even in minuscule quantities. George paused in the muggy Florida air to consider the implications: a readily available and cheap technology that could recognize threat liquids.
We called in the AT vendors, gave them our data on hydrogen peroxide detection and told them to start working on an algorithm that would automatically set off an alarm when hydrogen peroxide was detected. George insisted that the companies share their system performance and image format data so that TSA could open up algorithm development to a much wider audience of image analysis experts. That was strongly resisted by the manufacturers who wanted to retain their software rights and argued that they could move faster on their own. But, by mid-2009, George no longer had the air cover that I had provided, and the decision to let TSA share test data was reversed and each vendor is now free to bundle their proprietary black box scanners on their own time and price schedule.