UCSD/DARPA

After 33 days of painstaking puzzle-solving, a group of San Francisco-based computer programmers has un-shredded five pulverized documents to emerge as the winner of the  Defense Department’s Shredder Challenge .

The Defense Department’s research and development branch, the Defense Advanced Research Projects Agency, announced today that “All Your Shreds Belong to U.S.” had won the $50,000 challenge three days before the contest’s deadline.

Using custom-built computer software, the eight members of “All Your Shreds Belong to U.S.” reassembled seven pages of documents that had been shredded into more than 10,000 fragments in order to answer questions about the coded messages written on the pages.

“We are really stoked about it,” said Otavio Good, the winning team’s lead programmer. “From the start I said this isn’t about the money. It’s about getting exclusive, worldwide bragging rights.”

Close to 9,000 teams attempted the challenge, which was launched as a way for the Pentagon to find the fastest, most effective way to pull intelligence from shredded documents in war zones and also assess the possible security threats to America’s secret information.

“Lots of experts were skeptical that a solution could be produced at all let alone within the short time frame,” said Dan Kaufman, director, DARPA Information Innovation Office in a statement. “We are impressed by the ingenuity this type of competition elicits.”

Some teams, such as the one led by Manuel Cebrian at The University of California, San Diego, took a “crowd sourcing” approach to solving the puzzles, creating an online program where people from around the world could collectively work to solve what he called “one of the hardest puzzles ever proposed.”

Craig Landrum, co-founder of a document imaging group in Virginia, approached the problem solo, laboriously piecing each paper shred together as if solving a jigsaw puzzle.

But the winning strategy was all about the computer programming. Good, along with two other programmers, developed visual-recognition software that recommended possible matches when a user clicked on a particular paper fragment.

“Imagine if you’re playing a regular puzzle,” Good said, describing his custom-built program. “Pieces are scattered around. You click the place that you want to match a piece to and the computer recommends a number of pieces ordered by score and you chose which one you like the best.”

Good said it took about 600 hours to create the matching program and piece together the puzzles. About eight people were involved in the month-long effort.

“The major thing we did was program,” Good said. “Once we had the program, well, the puzzle pieces fell into place.”

But despite the “cool” new program’s success at solving the shredder challenge, Good said it would be much harder to reassemble cut-up documents in a real-world scenario. He said in most cases people should not be worried that their shredded secrets are going to get pieced back together because “it’s incredibly difficult.”