Re: [dm-crypt] Random fill

From: test532@codingninjas.org
To: dm-crypt@saout.de
Subject: Re: [dm-crypt] Random fill
Date: Mon, 31 Aug 2009 09:23:03 -0400	[thread overview]
Message-ID: <200908310923.03429.test532@codingninjas.org> (raw)
In-Reply-To: <874orow3rj.wl%htd@fancy-poultry.org>

> At Mon, 31 Aug 2009 06:50:39 -0400,
> test532@codingninjas.org wrote:
> 
> Thanks for your polite explaining so far!
> 
> > Because the point of filling with random data is to eliminate the
> > possibility of being able to tell where real data is stored.
> 
> Yes, that's clear.
> 
> > If the random data is cracked by using a known plaintext attack, then the
> > benefit of having this random data is nullified.
> 
> To crack the random data, the attacker must be able to distinguish them
>  from the "real" data. AS far as I know, if the key after a nullifiying
>  action via dmcrypt is wiped, the disk is filled with pseudorandom data and
>  ciphertext of the real data. For a known plaintext attack to work, one
>  must have the plaintext and the corresponding block of ciphertext
>  resulting in encryting this plaintext, ...

The truth in what you say above does make it difficult, but the amount of 
difficulty depends on how full the drive is. On a mostly empty drive, the 
probability of guessing correctly where the plain text equals zeros is fairly 
high. The further from mostly empty the drive becomes, the less likely success 
becomes.

> but that is not possible in this
>  scenario. No one can distinguish between the real ciphertext (encrypted by
>  a wholly different key) and the filling with zeroes.

> 
> The aim of a known plaintext attack is to find the key, but it has been
> irrevocably wiped, and the real text is encrypted by a totally different
>  key, in fact.
> 
> Sorry, but I can not see at all how a procedure which uses /dev/zero and
> dmcrypt to fill a partition with pseudorandom data could weaken the whole
> encryption / reduce the time for an attack on the real ciphertext at all.
> 

> Besides, AES is not vulnerable to known plaintext attacks, as far as I
>  know.

As far as you or I know. But I can guarantee you that someone out there knows 
more and isn't saying. How much more? I don't want to bet my data on it.

My operating system as it is right now has no vulnerabilities in it that I 
know of. That doesn't mean that I am going to turn off SELinux on the machine, 
because I know there is at least a possibility of vulnerabilities, and that is 
enough reason to take any extra precautions that are reasonable.

Sam

> 
> 
> _______________________________________________
> dm-crypt mailing list
> dm-crypt@saout.de
> http://www.saout.de/mailman/listinfo/dm-crypt
> 