Security Basics mailing list archives
Re: Wiping a drive: /dev/zero or /dev/urandom better?
From: Roman Fulop <ml () ensof1 trithem sk>
Date: Wed, 15 Oct 2008 11:35:51 +0200
AFAIK, /dev/urandom is pseudo-random generator, while /dev/random generates random data from entropy pool. /dev/random blocks if there is no entropy available, so i would not recommend it for overwriting disks. Roman Gaizka Isusquiza wrote:
Hi! Depending on the disk nature(i.e: optical support, magnetic support) several aproximations can be done. Supposing that we're talking about magnetic devices, we must know/remember some basics that can spoil all our efforts if fogotten. Is not the same to logically remove data than do it physically. Remove data logically means that using software means, there is no access to data, neither data can be rebuilt. The problem is that due to magnetism, magnetic devices have some sort of 'memory', which turns out to be main reason to rewrite device with senseless data several times(last time I read about this a minimum of 7 times). When trying to generate senseless data, best way in Linux is using /dev/urandom. /dev/urandom seeds are more unpredictable than /dev/random, mainly because /dev/urandom uses more unpredictable devices as input (i.e: analogic microphone) than mouse movement, mouse position on the screen or key pressed, which helps getting a good seed needed when generating a pseudorandom output. Remove data physically means damaging/destroying physical devices. What we must know is that there are means of recovering data from magnetic devices even when devices have been seriously physically damaged because there is a need to recover data after impredictible catastrophes(i.e: fires, floods). But this means are usually too expensive to be used by most crackers. At this moment, best way to prevent a disk to be rebuilt is using some kind of strong acid. But as I said, impossible is nothing, it's all a matter of money(if the value of the info is less than the money needed to get it, it's not worth trying to recover it). I hope I explained myself. Regards, Gaizka 2008/10/14 JW <jw () mailsw com>:I've got a theoretical question: when wiping a drive (I'm talking about Linux here), which of the following is more: fill the drive with data from /dev/zero or /dev/urandom? I ask because I often see people suggest something like the following for wiping disks: cat /dev/zero > /dev/hda (and of course do it multiple times) I got to thinking that (if you are really paranoid) it would probably be easier for "the bad guy" to recover original data if you use /dev/zero because it's so uniform, the "bad guy" can just look for anything other then zeros - if it's not zero, it's data. Which would imply that overwriting the data with /dev/urandom or /dev/random would be more secure. But I don't know enough about the internals of hard drives to know if it really matters or not. For clarity I'll point out that I'm not talking about wiping files in the filesystem, I'm talking about wiping whole disks - I guess you'd say "at the block level". What do the resident experts here think? JW -- ---------------------- System Administrator - Cedar Creek Software http://www.cedarcreeksoftware.com
Current thread:
- 51% can be enough Was: Wiping a drive, (continued)
- 51% can be enough Was: Wiping a drive Alexander Klimov (Oct 16)
- RE: 51% can be enough Was: Wiping a drive Olatunji Nowlin (Oct 16)
- RE: 51% can be enough Was: Wiping a drive Murda Mcloud (Oct 16)
- RE: 51% can be enough Was: Wiping a drive Alexander Klimov (Oct 20)
- Re: Wiping a drive: /dev/zero or /dev/urandom better? Razi Shaban (Oct 16)
- Re: Wiping a drive: /dev/zero or /dev/urandom better? Ansgar Wiechers (Oct 16)
- Re: Wiping a drive: /dev/zero or /dev/urandom better? Roman Fulop (Oct 15)
- Re: Wiping a drive: /dev/zero or /dev/urandom better? Eric Kollmann (Oct 15)
- Re: Wiping a drive: /dev/zero or /dev/urandom better? Yinka Adeosun (Oct 16)
- Re: Upptime report tools? Kevin Liang (Oct 16)
- RE: Upptime report tools? Lim, James (GTS Pac Rim) (Oct 16)
- Re: Upptime report tools? CJ (Oct 16)
- Webb statistics program Mattias Hemmingtsson (Oct 16)