![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Data erasure standards (not)
Feb. 26th, 2004 12:06 pmI regularly read people mention things like "DoD," "NSA" and "IEEE" "standards" for the secure wiping of data from hard drives. I recall a discussion on some list a few years back on how most of those "standards" are mythical, but can't find it now. [EDIT - found it! A BUGTRAQ post by Simple Nomad]. So, I did some digging on my own.
Sure, I've seen Peter Gutmann's paper and Simson Garfinkel's paper on the topic of data wiping.
While there ARE several US military and gov standards, NSA doesn't seem to publish one, and IEEE appears to have no data deletion standard. Indeed, even the IEEE paper on the topic (by Garfinkel) -- doesn't cite any such IEEE standard.
Yet documents citing mythical standards and misquoting the ones that DO exist abound.
Heck, one product blurb says the DoD's 5220.22-M recommends 7 overwrite passes. A look through the document shows it most strongly recommends burning or other utter destruction and has no mention of 7 (or any number of) passes. Interpretation of "destruction" is left to the branches.
It's possible some branch has interpreted it that way, but there is no DoD STANDARD on overwriting. Even Garfinkel seems to have pulled some ideas out of the DoD guidelines that don't actually exist in the documentation.
Some products claim to delete to Gutmann's "standard." Yet Gutmann talks about how his algorithm (which takes 35 passes) is more about him making the point that you have to know what you're erasing before you make a standard. That's in his paper's epilogue, which makes a great case for there not being a general standard.
So, on to the branches. The Army recommends heavy overwriting only if the drive is going to be used in an environment with the same or better security clearance, and recommends degaussing otherwise. Finally, the USAF guideline has the closest thing to an overwrite standard: triple overwrite: 0s, then 1s, then random, then verify. And then it's recommended only on a "case-by-case" basis, with disassembly and degaussing platters as the basic approach.
So, IS there a true public standard anyone has seen? I've just seen a lot of BS in advertising.
Sure, I've seen Peter Gutmann's paper and Simson Garfinkel's paper on the topic of data wiping.
While there ARE several US military and gov standards, NSA doesn't seem to publish one, and IEEE appears to have no data deletion standard. Indeed, even the IEEE paper on the topic (by Garfinkel) -- doesn't cite any such IEEE standard.
Yet documents citing mythical standards and misquoting the ones that DO exist abound.
Heck, one product blurb says the DoD's 5220.22-M recommends 7 overwrite passes. A look through the document shows it most strongly recommends burning or other utter destruction and has no mention of 7 (or any number of) passes. Interpretation of "destruction" is left to the branches.
It's possible some branch has interpreted it that way, but there is no DoD STANDARD on overwriting. Even Garfinkel seems to have pulled some ideas out of the DoD guidelines that don't actually exist in the documentation.
Some products claim to delete to Gutmann's "standard." Yet Gutmann talks about how his algorithm (which takes 35 passes) is more about him making the point that you have to know what you're erasing before you make a standard. That's in his paper's epilogue, which makes a great case for there not being a general standard.
So, on to the branches. The Army recommends heavy overwriting only if the drive is going to be used in an environment with the same or better security clearance, and recommends degaussing otherwise. Finally, the USAF guideline has the closest thing to an overwrite standard: triple overwrite: 0s, then 1s, then random, then verify. And then it's recommended only on a "case-by-case" basis, with disassembly and degaussing platters as the basic approach.
So, IS there a true public standard anyone has seen? I've just seen a lot of BS in advertising.
