At 2:01 PM +0200 9/14/99, Kuehn, Ulrich wrote:
>Hi,
>
>please find here included a mail from Holger Klawitter, the author of one of
>the mentioned palm cipher programs... He would have liked to recieve the
>critique himself.
>
>Ulrich
>
>-------------------------------------------
>[quoted form Holger Klawitter <[EMAIL PROTECTED]>]
>
> > > http://www.klawitter.de/palm/cipher.html uses IDEA to encrypt the
> > > clipboard, but it's ascii armouring would make it hard to manually
> > > transmit ciphertext, if I understand what he is doing.
>
>One of the major drawbacks of the Palm pilot is the maximum length
>of a note which is 4096 bytes. ASCII Armouring makes text longer so
>I decided to use an armour with 7bit significance. If the demand
>proves to be more base 64 encoding I'll be happy to switch.
Sorry, I was not critiquing Holger's program here, but rather
considering its suitability for an application very different from
its original purpose. The problem was how to enable a human rights
worker in the field to send and receive coded messages. Such a
worker might not have access to e-mail, so it is desirable that
messages be capable of transmission via snail-mail, FAX, voice, Morse
code, carved coconut shells or whatever. In this situation, even
base-64 would be a pain. In its original contest, 7/8 encoding makes
sense.
> > > Passphrase
> > > length is limited to 16 characters, which is unfortunate.
>
> > > Also, I wonder if the lack of a keyboard would make it a pain to
> > > enter persnickety text like passphrases and ciphertext.
>
>I think you've predicted my answer.
In general, I think forcing a user to pick a short password is a bad
idea. Even though entering a long password or passphrase on a Pilot
is awkward, that choice should be left to the user. A password
consisting of 16 random letters of the Roman alphabet offers 75 bits
of entropy. That is adequate for strong security, but barely. If a
user wants a strong passphrase that is easy to remember, 16
characters is hopeless.
The consensus recommendation for long term security is 90 bits. To
get that from 16 random characters, each character must be drawn from
an alphabet of 50 symbols.
>
>Side note to document what some people really (believe to) want:
>
>I got some request for storing the passord in a database in order
>not to have to type it in so often.
>
>Of course, I denied the request :-)
I agree that this is a dumb request if the program is only being used
to protect files on the Pilot. But I wouldn't dismiss the suggestion
quite so fast if the program is also being for communications
security. One possible solution might be to allow two passwords, one
of which is stored on the pilot in encrypted form, with the second
password as the key.
Arnold Reinhold
