Re: Playfair cracker - measure of best fit?
From: Douglas A. Gwyn (DAGwyn_at_null.net)
Date: 08/09/05
- Next message: David Wagner: "Re: Playfair cracker - measure of best fit?"
- Previous message: Regis: "Re: Should be in crypto for John E. Hadstate"
- In reply to: Rob Sullivan: "Playfair cracker - measure of best fit?"
- Next in thread: David Wagner: "Re: Playfair cracker - measure of best fit?"
- Reply: David Wagner: "Re: Playfair cracker - measure of best fit?"
- Reply: Joe Peschel: "Re: Playfair cracker - measure of best fit?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Tue, 9 Aug 2005 15:18:26 GMT
Rob Sullivan wrote:
> The problem I'm having is with step 2 - compared to other Playfair
> crackers, mine can only try a fraction of the keys tried by the other
> programs in a given time. Is there a better statistic I can use, in
> terms of appropriateness and efficiency?
Yes. Pearson's chi-square is unreliable when there are fewer
than, roughly, 5 entries per bin. Kullback's information measure
works all the way down to 0 entries, is easy to aggregate (add d.f.
and add information measures), and is easy to interpret (via
inverse chi-square lookup). You can obtain documentation and an
implementation I put into the public domain long ago at URL
http://www.phaedsys.demon.co.uk/chris/crypto/i-hat.zip
(If that link is broken, just Google for "i-hat" and "Kullback".)
- Next message: David Wagner: "Re: Playfair cracker - measure of best fit?"
- Previous message: Regis: "Re: Should be in crypto for John E. Hadstate"
- In reply to: Rob Sullivan: "Playfair cracker - measure of best fit?"
- Next in thread: David Wagner: "Re: Playfair cracker - measure of best fit?"
- Reply: David Wagner: "Re: Playfair cracker - measure of best fit?"
- Reply: Joe Peschel: "Re: Playfair cracker - measure of best fit?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|