In autumn 1989 confidential files from East Germany’s State Security Service (STASI) ought to have been destroyed almost overnight. The result : about 45 million torn documents are kept till today in 16250 bags.

On behalf of the BSTU, the institute which today deals with the former STASI files and which holds the bags with the shredded documents, the Fraunhofer Institute for Production Systems and Design Technology IPK now has the official task of putting together this gigantic puzzle. In the next two years, the content of initially 400 bags will be automatically reconstructed.

In order to guard the secrets of the GDR regime, the files of the Ministry for State Security were systematically destroyed between autumn 1989 and January 1990. The sheer number of files was so enormous that the shredders could not cope. A large amount of the files had to be torn by hand. Approximately 45 million DIN-A4 pages were torn into between eight and thirty pieces. Until now only a small amount of these documents has been successfully reconstructed, because the manual work is very time consuming.

In order to piece together the almost 600 million fragments about 30 persons would be occupied for 600 to 800 years. Researchers from Fraunhofer IPK can accomplish the task much faster : they have developed a software with which the puzzle can be solved automatically. Already in 2003 the feasibility of the underlying virtual puzzling technique has been demonstrated. Now the automatic reconstruction begins.

Before actually putting the pieces together each fragment has to be scanned from both sides and digitally stored. This process is done by the arvato direct services GmbH which belongs to Bertelsmann AG. Since 2005 arvato services and Fraunhofer IPK work together on digitalizing various documents especially for public administration and financial institutions.

During this cooperation new concepts for scanning have been developed which now are used in digitalizing the STASI fragments. »Virtual puzzling follows the logic of manual puzzling« , explains Fraunhofer IPK’s Dr. Bertram Nickolay, Head of Department of Security and Testing Technologies. To solve a puzzle is a test in patience. There are a lot of features in relation to which a human person decides whether two pieces fit or not : the outline of the pieces, their colour or even the pattern on their surface.

Such a pre-selection makes searching for and finding fitting puzzle pieces much easier. »The virtual puzzle process starts just in the same way« , says Nickolay. »The system analyzes several describing features, for example form and texture, in order to reduce the number of possibilities. The final reconstruction is then only attempted with the remaining smaller number of pieces.« The fragments are examined for possible matching outlines. If such fragments have been found they are pieced together to larger documents. Fragment by fragment and page by page the former STASI files are reconstructed.

Meanwhile researchers at Fraunhofer IPK have advanced from the technology as presented in 2003 : their algorithms are able to not only reconstruct manually torn documents but machine shredded ones, too. This is extremely difficult because machine torn documents all have the same shape, a feature that is fairly quintessential in solving normal puzzles. Regarding shredded documents the pattern of the script on the surface of the fragments has to be used for comparisons instead. Already, for Germany’s Tax Authority a bag full of shredded documents has been successfully reconstructed. Source : ipk.fraunhofer, May 2007