not too accurate in parts (this was a priority long before the spy plane drama but usually associated with key material); still interesting: --- http://www.physorg.com/news69416938.html "Fail-Safe Techniques Erase Magnetic Storage Media" ... "This is a very challenging problem," said Michael Knotts, a research scientist in the GTRI's Signature Technology Laboratory. "We had to verify that the data would be beyond all possible recovery even with unlimited budget and unlimited time. Commercial devices on the market for data erasure just couldn't fill the bill, because they were magnetically too weak, they were physically too large and heavy, or they didn't meet stringent air safety standards." ... Producing a magnetic field sufficient to destroy data patterns required the use of neodymium iron-boron magnets custom-designed for the project and special pole pieces made of esoteric cobalt alloys. The magnets, which weigh as much as 125 pounds, had to produce fields sufficient to penetrate metallic housings that surround some drives. "We developed models for magnetic circuits that we could run through optimization codes to design the best shape to get the field that we needed," Knotts said. "It takes quite a magnetic field to get through the steel enclosures on some of the drives. We are producing magnetic fields comparable to those used in magnetic resonance imaging equipment, so these are not your ordinary refrigerator magnets." ... "This was certainly an unusual project," he said. "It's not often that we get paid to crush equipment in presses, blow things up and set off fires in microwave ovens." ---end-cut--- i occasionally use one of those noisy AC powered degauss'ers on hard disks but have wondered if it would pass the "unlimited budget and unlimited time" test. i wonder what it would take to restrain a disk through an MRI machine without it impaling itself into the coils... (some of those horror stories are pretty wild) initializing a disk with entropy afterwards is also annoyingly time consuming for big disks. i'm concerned about just how usable you can make this process, perhaps turning it into a background batch process requiring multiple disks so one can be used for the live system while an empty disk is being wiped and randomized. i wonder when Peter Gutmann's article on hard disk encryption is going to be published (if not already?) :P