3 ms·
Which imparts more entropy to a disk, Gutmann or thermite?
This stems from a question offline. Does anyone have some data we can throw at this little Fermi problem? For example, how much energy does a disk head have to transmit to flip a bit? I nknow there are several levels, etc. Let's go with a 3.5" 500 GB SATA. Order of magnitude.
- _pius 18y agoI'm curious what the actual question is. Is it your headline or what you have in the description? As you probably know, the point of Gutmann isn't just to flip bits . . . it rewrites them like 35 times to reduce the magnetic remanence of the physical drive material to something too difficult for someone to measure with precision unless they've got extremely expensive tools. If I were going Fermi on the "how much energy to flip a bit" question, I'd reason that the average desktop computer has a 400W power supply and that a hard drive would need to draw at least two orders of magnitude less power than the power supply in order to allow for a stable machine. So let's say the drive draws ~1 watt of power. I'd assume that the SATA drive is rated for 10,000 RPM. So let's say our 1 watt drive can hit ~100 drive locations per second. Let's assume that each drive location corresponds to a bit. A watt is just Joules/sec, so it takes one Joule to flip ~10^2 bits, or ~10^-2 Joules to flip a bit. I'd probably take that down an order of magnitude because I assume that the vast majority of the energy goes into moving physical components of the disk rather than into the actual charge imparted to flip the bits. So my answer would be ~10^-3 Joules to flip a bit.
- niels_olson 18y agocool. Yeah, I'm assuming whoever wants the drive is prepared to use electron microscopes and such to uncover the bits on the mangled wreckage. So the question isn't just a informational entropy question, but a physical entropy question. How much physical entropy does Gutmann 35 pass (or n passes) add to the disk? Basically, somebody asked a listserv what was the best way to securely wipe a hard drive. I said Gutmann 35 pass, and others came back with ideas for thermite, shooting it, etc. I figured, compared to Gutmann, thermite (the most exothermic chemical reaction, and one that happens to use iron) would be the closest. But I got to thinking, you know, if that thermite just blasts, then you've still got fragments and Gutmann wins. If it's a smoldering sort of thing and you have plenty of oxygen, maybe thermite wins, but how much thermite would you need, and how would you have to do it? So I guess a better question would be what's the minimum necessary chemical reaction to equal the entropy of a 35 pass Gutmann?
- _pius 18y agoIn practice, the best way to handle a drive that might actually need to be resistant to that level of analysis is to degauss the drive, shred it, and then melt it down in an incinerator. You wouldn't mess around with Gutmann at all because it's just a poor man's degauss, used for situations when you don't want to risk damage to the drive. You wouldn't just detonate the drive either because, as you mentioned, individual pieces could potentially still have recoverable data.