You Can Put Humpty Dumpty Together Again -- If He's a Shredded Document

May 21, 2007

Hard-working algorithms are at the heart of an $8.5 million software system that can piece together manually shredded documents.

According to Nature, the one-of-a kind document-reconstruction system was developed by researchers at Berlin's Fraunhofer Institute for Production Systems and Design Technology (IPK). The system scans the torn material, creates a digital image of each piece, and groups the images into clusters according to similarities in shape, texture, typeface, color, and so on. The system then fills in the missing information in the clusters in the same way that someone would solve a jigsaw puzzle.

The software "learns as it processes," said Bertram Nickolay, IPK's head of security technology.

The point of this system, which will begin operations in June, is to reassemble some 45 million pages—600 million pieces of shredded paper—left behind by East Germany's State Security Service (Stasi) in 1989-1990. As the East German government collapsed, agents tried to destroy evidence, leaving behind a mountain of shredded files.

A two-year pilot project will tackle just 2 percent of the documents, after which the government will decide whether to finance the rest of the unravelling. Suitably modified, the software could be used to reconstruct machine-shredded documents.

Source: Nature, May 10, 2007