San Francisco Team Wins DARPA's De-Shredding Contest 94
Posted
by
timothy
from the california-v-greenwood-1988 dept.
from the california-v-greenwood-1988 dept.
New submitter karlnyberg writes with an update to the recently announced de-shredding challenge posted by DARPA: "The team 'All Your Shreds Are Belong To U.S.' has correctly solved all five puzzles, and the Challenge has now ended. You may view the winning team's submissions as well as the complete puzzle solutions by following the links on our homepage. We recognize that many of our participants have devoted countless hours to painstakingly piecing our puzzles back together, and we truly appreciate everyone's efforts. Hopefully you enjoyed the Challenge and learned something new along the way. We certainly did!"
Re:As I said before... (Score:5, Informative)
You might be suprised. A fire can be dangerous, a shredder is convenient. Also, we can reconstruct documents that have been burned, and if you have a big stack it can actually take longer to burn them beyond recovery. Why? It's about oxygen availability - the corners, top and bottom pages will burn first, but the center of a pile of documents will often remain intact, yielding valuable intelligence if recovered.
Having looked up the shred, it seems to be standard commercial shred sizes - the DOD goes quite a bit smaller than that.
simplified test (Score:5, Informative)
The test was actually much simpler than any real-world application might be. Each puzzle was really only one or two (or a few) shredded pages, with various degrees of shredding and various bits of writing. It is a first step, but nowhere near what you would be dealing with in any real-world situation where hundreds or thousands of pages of shredded documents would be mixed together.
I participated (a bit) with the UCSD team that basically made a crowd-sourced jigsaw puzzle to do it - at last check they were in the top 5, but I don't think they got the last puzzle (yet). This approach seems reasonable for the relatively simple puzzles of the challenge, but it really wouldn't scale very well - requires a lot of labor.
It sounds like the winning team had a much better (and more scaleable) strategy, where an algorithm scores all of the pieces for fit in a particular place and then allows the user(s) to choose the best piece from a few high-scoring ones. While I still don't think this would work very well in a real-world scenario, obviously it would work better than depending on massive crowd sourcing.
Shredders are good enough for classified docs (Score:5, Informative)
If that's not good enough, some locations use burn-boxes - never trust a machine to do thermodynamic's job!
Re:Shredders are good enough for classified docs (Score:2, Informative)
The NSA-approved shredder I used made chad so fine that the frayed fibres
at the edges made up a large percentage and maybe the majority of the surface
of each chad. There is so much information lost in that area that it may not be
possible to reconstruct such documents in the future even with state-level resources.
Re:WTF? (Score:4, Informative)
Actually they weren't allowed to under previous law. So they simply paid the British and Australians to do it for them and did likewise for them.
Now that law doesn't exist so they don't need to bother with the loophole of paying a friendly country to do the spying job for them. You don't think the massive data center being built at Fort Williams in Utah is for spying on other countries do you?