🚀 go-pugleaf

[ The ">>" text is from knight@gnu.ai.mit.edu ]

In article <1992Aug13.181102.11451@cactus.org> ritter@cactus.org (Terry Ritter) writes:
>>     The increase is the excess blocking.  I encrypt in exactly
>>1024 byte (1K) blocks.  . . .
>> . . . due to computer file clusters, this block doesn't
>>take up any additional space from the drive.
>
> Well, we must be clear about whether the cipher is working on actual
> disk sectors or files.  Only files have been mentioned.
>
> I think it is important to end up (after enciphering and deciphering)
> with an exact duplicate of the original file, in data and length.
> Any change from variable-length (files) to block-oriented processing
> must add something to the original data to be able to recover the
> original length.  And if the original length was 1023 mod 1024, there
> will be no free space in the last block to hold additional data.
> Thus, any block-oriented cipher which works on variable-length files
> must expand some files.  Small blocks imply a small expansion, but
> frequently; large blocks imply a larger expansion less frequently.


You should both read more about DES's modes of use.

Ciphertext-stealing is an excellent way to encipher the short block
(last block) of a file without increasing the length.

A variant of this can be used for number-theoretic based cryptosystems
such as RSA to limit the size increase (of a file) to a single bit.

Also, blocksize of a filesystem is device-dependent.  It ranges from
512 bytes to 8K on ``typical'' Unix boxes.

Now back to the regularly scheduled junk.  And back to real work.

-bsy
-- 
Bennet S. Yee		Phone: +1 412 268-7571		Email: bsy+@cs.cmu.edu
School of Computer Science, Carnegie Mellon, Pittsburgh, PA 15213-3890