[dm-crypt] double algorithm question

[dm-crypt] double algorithm question

[Sam]

Hi All,

I am wondering if this is a good idea:

encrypt a partition normally with cryptsetup luksFormat (using aes-xts-plain),
then luksOpen,
mkfs.ext2 format the device mapper device that appears,
mount it.
Then, create a giant file that fills up the partition.
losetup it that file,
luksFormat the loop device (using twofish-xts-plain)
luksOpen it,
mkfs.ext2 format the device mapper device that appears,
mount it,
and use it...

My purpose is that I don't trust AES, but I don't trust twofish enough to be 
sure it is better than AES.

I am paranoid enough that the speed hit is acceptable.

Questions:

1) is this the best way to achieve my goal with dm-crypt?
2) is it secure? Or will somehow it cause my data to be less secure than just 
using one cipher? Or will it somehow defeat the security provided by XTS? (i 
would assume it becoming less secure in any way is impossible, but i am not a 
cryptoanalyst, so i don't want to be assuming such things).

I know truecrypt has a feature where you specify the cipher as aes-twofish. 
This is what I wish to achieve, but using dm-crypt.

Regards,
Sam

Re: [dm-crypt] double algorithm question

[Moji]

You do not need to make a filesystem on the intermediate device, because you treat the devices in /dev/mapper as block devices you can luksFormat any device that shows up in order to do cascade encryption. You just have to remember to close them first in last out.

cryptsetup luksFormat -c aes-xts-plain /dev/sdc
cryptsetup luksOpen /dev/sdc first_layer
cryptsetup luksFormat -c aes-xts-plain /dev/mapper/first_layer
cryptsetup luksOpen /dev/mapper/first second_layer
mkfs.ext2 /dev/mapper/second_layer -m 0 -L "Test"
mount /dev/mapper/second_layer /mnt/usb
umount /mnt/cdrom
cryptsetup luksClose second_layer
cryptsetup luksClose first_layer

[Of course omit the luksFormat/mkfs lines after the device is created to open/close the device.]

I do not know of any vulnerabilities with cascade encryption, it is normally just excessive, but someone else might.

I hope that helps you,

-MJ

On Sat, 1 Aug 2009 07:39:42 -0400
Sam <test532@codingninjas.org> wrote:

> Hi All,
> 
> I am wondering if this is a good idea:
> 
> encrypt a partition normally with cryptsetup luksFormat (using aes-xts-plain),
> then luksOpen,
> mkfs.ext2 format the device mapper device that appears,
> mount it.
> Then, create a giant file that fills up the partition.
> losetup it that file,
> luksFormat the loop device (using twofish-xts-plain)
> luksOpen it,
> mkfs.ext2 format the device mapper device that appears,
> mount it,
> and use it...
> 
> My purpose is that I don't trust AES, but I don't trust twofish enough to be 
> sure it is better than AES.
> 
> I am paranoid enough that the speed hit is acceptable.
> 
> Questions:
> 
> 1) is this the best way to achieve my goal with dm-crypt?
> 2) is it secure? Or will somehow it cause my data to be less secure than just 
> using one cipher? Or will it somehow defeat the security provided by XTS? (i 
> would assume it becoming less secure in any way is impossible, but i am not a 
> cryptoanalyst, so i don't want to be assuming such things).
> 
> I know truecrypt has a feature where you specify the cipher as aes-twofish. 
> This is what I wish to achieve, but using dm-crypt.
> 
> Regards,
> Sam
> _______________________________________________
> dm-crypt mailing list
> dm-crypt@saout.de
> http://www.saout.de/mailman/listinfo/dm-crypt

Re: [dm-crypt] double algorithm question

[Sam]

Thanks Moji,

That will obviously provide a nice boost in performance over what I was 
trying! I appreciate your help.

Regards,
Sam

> You do not need to make a filesystem on the intermediate device, because
> you treat the devices in /dev/mapper as block devices you can luksFormat
> any device that shows up in order to do cascade encryption. You just have
> to remember to close them first in last out.
>
> cryptsetup luksFormat -c aes-xts-plain /dev/sdc
> cryptsetup luksOpen /dev/sdc first_layer
> cryptsetup luksFormat -c aes-xts-plain /dev/mapper/first_layer
> cryptsetup luksOpen /dev/mapper/first second_layer
> mkfs.ext2 /dev/mapper/second_layer -m 0 -L "Test"
> mount /dev/mapper/second_layer /mnt/usb
> umount /mnt/cdrom
> cryptsetup luksClose second_layer
> cryptsetup luksClose first_layer
>
> [Of course omit the luksFormat/mkfs lines after the device is created to
> open/close the device.]
>
> I do not know of any vulnerabilities with cascade encryption, it is
> normally just excessive, but someone else might.
>
> I hope that helps you,
>
> -MJ
>
> On Sat, 1 Aug 2009 07:39:42 -0400
>
> Sam <test532@codingninjas.org> wrote:
> > Hi All,
> >
> > I am wondering if this is a good idea:
> >
> > encrypt a partition normally with cryptsetup luksFormat (using
> > aes-xts-plain), then luksOpen,
> > mkfs.ext2 format the device mapper device that appears,
> > mount it.
> > Then, create a giant file that fills up the partition.
> > losetup it that file,
> > luksFormat the loop device (using twofish-xts-plain)
> > luksOpen it,
> > mkfs.ext2 format the device mapper device that appears,
> > mount it,
> > and use it...
> >
> > My purpose is that I don't trust AES, but I don't trust twofish enough to
> > be sure it is better than AES.
> >
> > I am paranoid enough that the speed hit is acceptable.
> >
> > Questions:
> >
> > 1) is this the best way to achieve my goal with dm-crypt?
> > 2) is it secure? Or will somehow it cause my data to be less secure than
> > just using one cipher? Or will it somehow defeat the security provided by
> > XTS? (i would assume it becoming less secure in any way is impossible,
> > but i am not a cryptoanalyst, so i don't want to be assuming such
> > things).
> >
> > I know truecrypt has a feature where you specify the cipher as
> > aes-twofish. This is what I wish to achieve, but using dm-crypt.
> >
> > Regards,
> > Sam
> > _______________________________________________
> > dm-crypt mailing list
> > dm-crypt@saout.de
> > http://www.saout.de/mailman/listinfo/dm-crypt
>
> _______________________________________________
> dm-crypt mailing list
> dm-crypt@saout.de
> http://www.saout.de/mailman/listinfo/dm-crypt

Re: [dm-crypt] double algorithm question

[Roscoe]

FWIW I wouldn't forget about Serpent, which came 2nd in the AES
selection process.

From Practical Cryptography, by Niels Ferguson and Bruce Schneier:
"If  you are paranoid about the security of your data, and speed is
not important, then you should choose Seprent. During the AES process,
every serious cryptographer agreed that Serpent was the most secure
(or most conservative) of the submissions"

-- Roscoe

On Sun, Aug 2, 2009 at 12:48 AM, Sam<test532@codingninjas.org> wrote:
> Thanks Moji,
>
> That will obviously provide a nice boost in performance over what I was
> trying! I appreciate your help.
>
> Regards,
> Sam
>
>> You do not need to make a filesystem on the intermediate device, because
>> you treat the devices in /dev/mapper as block devices you can luksFormat
>> any device that shows up in order to do cascade encryption. You just have
>> to remember to close them first in last out.
>>
>> cryptsetup luksFormat -c aes-xts-plain /dev/sdc
>> cryptsetup luksOpen /dev/sdc first_layer
>> cryptsetup luksFormat -c aes-xts-plain /dev/mapper/first_layer
>> cryptsetup luksOpen /dev/mapper/first second_layer
>> mkfs.ext2 /dev/mapper/second_layer -m 0 -L "Test"
>> mount /dev/mapper/second_layer /mnt/usb
>> umount /mnt/cdrom
>> cryptsetup luksClose second_layer
>> cryptsetup luksClose first_layer
>>
>> [Of course omit the luksFormat/mkfs lines after the device is created to
>> open/close the device.]
>>
>> I do not know of any vulnerabilities with cascade encryption, it is
>> normally just excessive, but someone else might.
>>
>> I hope that helps you,
>>
>> -MJ
>>
>> On Sat, 1 Aug 2009 07:39:42 -0400
>>
>> Sam <test532@codingninjas.org> wrote:
>> > Hi All,
>> >
>> > I am wondering if this is a good idea:
>> >
>> > encrypt a partition normally with cryptsetup luksFormat (using
>> > aes-xts-plain), then luksOpen,
>> > mkfs.ext2 format the device mapper device that appears,
>> > mount it.
>> > Then, create a giant file that fills up the partition.
>> > losetup it that file,
>> > luksFormat the loop device (using twofish-xts-plain)
>> > luksOpen it,
>> > mkfs.ext2 format the device mapper device that appears,
>> > mount it,
>> > and use it...
>> >
>> > My purpose is that I don't trust AES, but I don't trust twofish enough to
>> > be sure it is better than AES.
>> >
>> > I am paranoid enough that the speed hit is acceptable.
>> >
>> > Questions:
>> >
>> > 1) is this the best way to achieve my goal with dm-crypt?
>> > 2) is it secure? Or will somehow it cause my data to be less secure than
>> > just using one cipher? Or will it somehow defeat the security provided by
>> > XTS? (i would assume it becoming less secure in any way is impossible,
>> > but i am not a cryptoanalyst, so i don't want to be assuming such
>> > things).
>> >
>> > I know truecrypt has a feature where you specify the cipher as
>> > aes-twofish. This is what I wish to achieve, but using dm-crypt.
>> >
>> > Regards,
>> > Sam
>> > _______________________________________________
>> > dm-crypt mailing list
>> > dm-crypt@saout.de
>> > http://www.saout.de/mailman/listinfo/dm-crypt
>>
>> _______________________________________________
>> dm-crypt mailing list
>> dm-crypt@saout.de
>> http://www.saout.de/mailman/listinfo/dm-crypt
> _______________________________________________
> dm-crypt mailing list
> dm-crypt@saout.de
> http://www.saout.de/mailman/listinfo/dm-crypt
>

Re: [dm-crypt] double algorithm question

[Roscoe]

[Post attempt two, first one got blocked]

FWIW I wouldn't forget about Serpent, which came 2nd in the AES
selection process.

From Practical Cryptography, by Niels Ferguson and Bruce Schneier:
"If  you are paranoid about the security of your data, and speed is
not important, then you should choose Seprent. During the AES process,
every serious cryptographer agreed that Serpent was the most secure
(or most conservative) of the submissions"

 -- Roscoe