RE: Double Encryption Is Not A Good Idea

["Owen Lewis" <[EMAIL PROTECTED]>]

> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf Of Dave Howe
> Sent: 30 October 2004 15:41
> To: [EMAIL PROTECTED]
> Subject: Re: Double Encryption Is Not A Good Idea
>
>
> Brian Beesley wrote:
> > Owen, your scheme seems to be fine for what it's intended to
> do, but, for
> > lossless compression of digital data, it seems to be either
> snake oil or
> > black magic.
> For that matter - it would represent a breakthough in breaking cyphers.
> given the example is an instant solution for a chosen-plaintext attack
> against des (a lookup table), you could precompute only a fraction of
> the data space, then feed those values into owen's equation for some
> other, unknown cryptotext value and out would pop the key....

'Fraid not. Not ever.

The question was whether or not it could effect appreciable savings in the
storage space of n strings of data, all of a common length.

To do this, within the storage space needs to be held information on the
value of each byte in a string and the position of each byte in the string
to be able to re-create the string perfectly.

The method can't supply a satisfactory answer because, to make a random data
string an analogy for activity in the RF spectrum requires that each byte in
the string be considered analogous to a discrete and active RF channel.
Though arranging the data string in a set pattern that makes it 'digestible'
to the algorithm seems trivial and the information on discrete byte position
in the original string can be thus markedly compressed, this cannot not give
an affirmative answer the question because:

	-	The information of the value (amplitude) of each byte is incompressible.
Therefore, storing the requisite information using the algorithm proposed
might actually require an *increase* of storage space over and above that
required to make simple copy of the original input data. Hardly efficient
for the purpose :-(

	-	The technique described and as currently applied involves the search of a
large space for a relatively small about of information. Whether the
information is random in either or both of amplitude (value) or distribution
is irrelevant, since there is no exploitation of any pattern that might
exist. The key to its effective use is that the volume of original data is:

			a. Much smaller than the search space. This limitation is definitive.

			b. Smaller than the number of bytes in the truncated harmonic analogue of
the original data that is used as a method of storing a description of the
original data set.

So, whilst we get an excellent compressive effect for our purpose, in we can
store all information on activity above a threshold level and in a data
string typically 2.5% only of the size required to make a copy of the
original *search space*(as is done in some other methods of spectrum
search), there is no general application to the storage of ascii data
strings. Such strings can only be made analogous to activity in the RF
spectrum, in so far as 'space' between the bytes is first artificially
created, I.e. the data is left unaltered, as to each byte's value and
ordering, but an artificial search space is created within which to
re-present it.  This technique is ultimately self defeating since the
'compressed' data record only will achieve a compression of the artificially
created search space and not of the source data itself.

So.... a nice compressive technique for a special purpose but without
application to the compression of ascii strings. The only application seems
to be the recording in amplitude and placement of a relatively small amount
of data within a very large search space - and that it can do well. But no
Phlogiston.

Owen