Slashdot Mirror
Single Drive Wipe Protects Data
ALF-nl writes "A forensics expert claims that wiping your hard drives with just one pass already makes it next to impossible to recover the data with an electron microscope." But that's not accounting for the super secret machines that the government has, man.
Hard drive meets Mr. Thermite.>
Actually you can...
If each bit is stored with multiple pins and the majority of the collection are in the 1 position but a few of them are in the 0, you would just take the dominant state.
That is assuming that the dominant majority should be correct. Perhaps the average of the bits should be used? This example is hardly fair and all... damned computer science is pushing it's winner-takes-all political views into my hardware!
Perception is the thin dividing line between reality and fiction.
Wow, they put the prize money up! Last time we discussed that here, the prize was a whopping $40.
By reading this signature, you hereby agree with the content of the above comment.
Why not:
dd if=/dev/random of=/dev/hda
instead?
That way you get random data, not just all zeros. Also you probably want /dev/hda so you blank the entire drive; not /dev/hda1 which only blanks the first partition.
Cheers,
Dave
They that can give up essential liberty to obtain a little temporary safety deserve neither safety nor liberty.
Ben
Add a wipe to the encryption and you may be safe.
The old problem with multiple wipes depended on the fact that there were rather large tolerances, but modern drives are very close to limits caused by physics, which means that it's a lot harder to extract wiped data.
If the data also was encrypted it will probably be impossible to re-create since there always is a level of loss even at recovery. For unencrypted data this may not be a big problem and it can be rectified by hand, but for encrypted data it will upset the whole packet that was encrypted.
But in a majority of cases a single wipe will be sufficient when the hardware is sold as surplus, since it's not easy to track and find out if a certain drive contains anything of interest.
If builders built buildings the way programmers wrote programs, then the first woodpecker would destroy civilization.
That would take too long - you can't depend on the blocking kernel random generator, as it needs a source of data to keep feeding the entropy pool.
I want to delete my account but Slashdot doesn't allow it.
That'd probably be this challenge from further up the page - $500 at the moment, and apparently three companies have turned it down after the dd command was mentioned because they 'know' it isn't possible.
Note to the clueless: The above comment is entirely intended to make the point that encryption is not a substitute for wiping. If you can recover encrypted data with a key, so can someone who doesn't have the key given enough time, skill and determination. It's not just a theoretical possibility.
By showing myself to be sounding confident with an obviously wrong statement, I was parroting the parent.
My blog
The source of the claim seems Gutmann's 1996 article: http://www.usenix.org/publications/library/proceedings/sec96/full_papers/gutmann/index.html where he says: "Data overwritten once or twice may be recovered by subtracting what is expected to be read from a storage location from what is actually read. Data which is overwritten an arbitrarily large number of times can still be recovered provided that the new data isn't written to the same location as the original data (for magnetic media), or that the recovery attempt is carried out fairly soon after the new data was written (for RAM)." It was challenged already in 2003 http://www.nber.org/sys-admin/overwritten-data-guttman.html where Feenberg writes: "Surveying all the references, I conclude that Gutmann's claim belongs in the category of urban legend." As usual, this story shows that individual claims have to be checked by independent parties. Even the claim that it can not be done.
You want /dev/urandom. Pseudorandom data is plenty for this purpose, and it won't take forever to generate either.
Give me Classic Slashdot or give me death!
Yes, a two year old with Torx for fingers. But seriously if you really want to know how to erase your media here are the instructions for the US government http://csrc.nist.gov/publications/nistpubs/800-88/NISTSP800-88_rev1.pdf. For destroying hard drives they recommend you "disintegrate, shred, pulverize, incinerate" (p19) the hard drive
Tanto nomini nullum par elogium.
Under normal conditions /dev/random would likely take decades, if not centuries, to do the wipe.
Gutmann 35-pass is designed for hard drives which use MFM/RLL encoding. New disks don't use this encoding anymore, so this method is pretty much equal in deletion quality to the other methods.
Well, the feds only specify that unclassified drives be wiped. Classified drives (that is, hard drives with classified material on them) must be destroyed.
Incidentally, a lot of data on hard drives is user-inaccessible due to the hard drive remapping bad sectors. Only a low-level format will touch that data--not application-level wiping tools.
> if you can recover from 1 overwrite, while still being able to get the new data, the
> capacity has just doubled.
Not if it takes hundreds of hours to do and recovers only 3/4 of the data on average. There is a lot of room between "not secure" and "reliable data storage".
It is very unlikely that any of us need worry that our overwritten files will be recovered, though. None of us have secrets that important.
Besides, the bot that controls your Windows box has already uploaded all your passwords.
Warning: this article may contain humor, sarcasm, parody, and perhaps even irony. Read at your own risk.
Last time I cared about government standards for this sort of thing, I got the NSA document describing standards for the government. It basically reduced down to "take the hard drive platters, and grind them to dust".
While I may doubt the government in general, if NSA says wiping isn't sufficient, I'm inclined to agree with them.
"I do not agree with what you say, but I will defend to the death your right to say it"
There is *no* way to recover the data on a modern drive after a single wipe. It is actually impossible. It cannot be done.
The reason is simple - although you may be able to detect a tiny tiny bit of data from the previous recording, you've no idea how strongly overwritten it is. Now, with old drives which used simple on/off pulses to write data to the disk, it would be possible to see if the bit you're looking at is a little higher or lower than it should be, and infer the previous value from that. Modern drives use a system similar to QAM - quadrature amplitude modulation - to pack more bits of data into each transition on the disk. Since the signal is essentially analogue, you'd need to know how badly degraded the print-through was. You can't do this, so you can't recover data after it's been overwritten even once.
Pop in a DBAN cd, hit enter. You can tell the boss that you've performed a wipe that meets DoD specifications. There's no real time difference in doing one wipe, which doesn't meet DoD specs, or the three that DBAN does by default. Unless, of course, you are sitting there watching the percent complete go up. If you have free time to do that, how can I apply for your job?
For the google impaired, http://www.dban.org/
Evidence of what?
You know it is often important to hide data that isn't involved with anything illegal. For example: Credit Card numbers, social security numbers and other personal information, trade secrets, personal journals and diaries that you don't want other people reading. There are many MANY reasons to want to wipe data that doesn't implicate you in a crime.
In the epilogue of http://www.cs.auckland.ac.nz/~pgut001/pubs/secure_del.html, Peter Gutmann basically calls the author of TFA a rtrd.
Apparently, he's confusing two different techniques, and Gutmann claims that, of course it won't work the way he's doing it. He's doing it wrong. You can't use the Magnetic Force Microscope to perform an error cancelling read, it doesn't work. The success rate is - surprise! - less than 1%, exactly like TFA claims.
Also, mentioned in Gutmann's epilogue, TFA confuses an MFM and a scanning electron microscope. They are not the same thing. An MFM reads magnectic levels, it doesn't "see" electrons like a SEL will.
In any case, Gutmann agrees with TFA but for very different reasons. The new encoding techniques nullify the MFM. There is no point using it because it won't give you any usefull information on a modern drive. Also, the extremely high densities mean the only practical and reliable method of recovery is basic error-cancelling techniques, and that's only practical after one wipe. Even then, it's iffy at best.
So yes, a single wipe is probably all you need. But who knows what data recovery techniques will be invented? A single pass is probably good enough right now, but 3-4 random passes is pretty much a sure thing, regardless of future techniques.
Security is mostly a superstition... Avoiding danger is no safer in the long run than outright exposure. - Helen Keller
You are wrong. Because any decent hard drive encryption solution will not use the password to directly encrypt the sectors. They will use it to encrypt one ore more master keys which will then be used to encrypt sectors. For example dm-crypt/LUKS works that way (up to 7 master keys), as well as TrueCrypt. They do that precisely to render all the data inaccessible by simply wiping the master key. Another advantage of this technique is that the user can change her password at anytime without having to re-encrypt the whole disk (the app just re-encrypts the master key).
So the GP is right: use disk encryption instead of relying on time-wasting/manual/unreliable data wiping !
That would take too long - you can't depend on the blocking kernel random generator, as it needs a source of data to keep feeding the entropy pool.
True. Grandparent probably meant:
dd if=/dev/urandom of=/dev/hda
(For people who aren't UNIX geeks: /dev/random gives you really good random numbers, the sort you'd want if you were generating a PGP key. If it runs out of random numbers then it blocks, and doesn't return until something happens to give it a good random number (e.g. an interrupt, where the nanosecond timing is pretty random). /dev/urandom gives you numbers that are random enough for most purposes, and it doesn't block).
One reason they require it is simple paranoia. The lengths you go to protect something depends on the value of the thing you are protecting and thus the lengths someone might go to get it. Same reason they use lots of armed, highly trained agents to protect the president. The president is extremely important to the nation and people will go to great lengths to harm him. When you are talking about classified data, you go to the paranoid extreme.
Another reason is inertia. These rules were written back when drives were much simpler and thus easier to recover data from. However the government moves slow and hasn't bothered to update. Remember that time was disks used frequency modulation to store their data. It was a pure binary "every thing above this level is a 1 everything below this other level is a zero." Thus it was much easier to infer what the previous data had been. Now drives store an analogue waveform and analyze that to determine the maximum likely data it represents. It's call EPRML. It sounds like voodoo, but works great and is very reliable. It also plays hell with any attempt to figure out what was on there before since there are no fixed levels for 1 and 0.
So I'm not saying don't do multiple wipes. It doesn't hurt, just realize that just because the government does it doesn't mean you need to do it too. Remember that one wipe screws over any and all methods that don't involve disassembling the drive. So unless you think someone is so interested in your data they'll take the drive apart and put it under a microscope, then one wipe is all you need. That is a whole shitload of work, and requires rather specialized equipment and training. You worried about people like that after your data? You think if they were that interested they wouldn't maybe just come and put a gun to your head to get it?
You need to wipe your drive because it's easy for any bozo to run a program that looks at what's in unallocated space. However you only need one wipe to prevent that.
According to Bruce Schneier:
We've never factored a 1024-bit number -- at least, not outside any secret government agency -- and it's likely to require a lot more than 15 million computer years of work.
So even if the usable computational speed of processors doubles in the next few years, it would still take at least 7.5 million computer years of work. You might have that much time (or maybe you have 7 million computers) but I don't.
No, increased computational speeds won't make factoring extremely large numbers feasible (at least, not anytime soon). The only thing that will do that would be finding some algorithm to do it - and if you figure that out, you'll deserve every award you get and then some.
Well, as MC Frontalot says, "You can't hide secrets from the future with math."