Bit Balancing (or not) by table lookup

From: John E. Hadstate (nospam@null.nil)
Date: 03/27/03

  • Next message: Mads Rasmussen: "Re: SHA-1 on the web (Java Applet)?" 
    From: "John E. Hadstate" <nospam@null.nil>
Date: Thu, 27 Mar 2003 06:24:07 -0500

    (1) Create a table of the 12,870 unique 16-bit words that have exactly 8
    bits equal to 1.

    (2) From this table, randomly draw 256 words and store them in a second
    table.

    (3) Lookup incoming plaintext bytes in the table and replace the plaintext
    bytes with 16-bit, bit-balanced, words.

    The obvious advantage is high speed (not considering the potentially slow
    setup time), assuming you have some reason for bit-balancing. The obvious
    disadvantages are inefficient use of bandwidth and the computational cost of
    inverting step (3).

    Now, consider this variation. In step (1), instead of creating a table of
    the 12,870 combinations of 16 bits taken 8 at a time, create a table of the
    8,008 combinations of 16 bits taken 6 at a time and use it for input to step
    (2).

    I claim that even though the output stream will be heavily unbalanced in
    favor of 0's (5/8 0's, 3/8 1's) it will be impossible to exploit this fact
    because every output word shows exactly the same statistics.

  • Next message: Mads Rasmussen: "Re: SHA-1 on the web (Java Applet)?"