From: futplex@pseudonym.com (Futplex)
Date: Fri, 20 Oct 95 21:01:16 PDT
To: privsoft@ix.netcom.com
Subject: Re: [reformatted] how secure can privasoft be?
In-Reply-To: <199510202040.QAA29119@light.lightlink.com>
Message-ID: <199510210401.AAA16706@opine.cs.umass.edu>
MIME-Version: 1.0
Content-Type: text/plain


> PrivaSoft has obtained an export license from the governments of Israel and 
> the USA.  
[...]
> PrivaSoft willfully complies with these regulations as it is a commercial 
> product, and it is not intended for national security applications with its 
> current key length which is the maximum legally allowable for commercial 
> users.
> 
> The cryptographic engine of PrivaSoft
> 
> PrivaSoft uses a pseudo-random generator that is seeded by a 9 digit number 
> uniformly normalized from the user's secret key.  The engine is proprietary, 
> designed according to the rules of modern cryptology to make the best use of 
> the allowable key length.

This seems paradoxical. PrivaSoft uses a key approximately 30 bits long. It
is claimed that the key length "is the maximum legally allowable for 
commercial users". How does this square with (to pick a familiar example) the
40 bit RC4 keys used in the exportable international version of Netscape 
Navigator ?

Two possible explanations I can imagine are:

(0) PrivaSoft actually uses a key longer than 9 digits, and someone just made
    a mistake somewhere along the line. Perhaps it's really 9 extended-ASCII
    characters == 9 bytes == 72 bits, rather than 9 digits == 
    ceil(9*lg(10)) bits == 30 bits.

(1) NSA believes it is about (2^40)/(2^30) == 1024 times harder to break
    the PrivaSoft scheme than SSL with 40-bit RC4. That would be remarkable.

[...]
> The use of default keys
> 
> When secret keys or passwords are used by laymen, there is always a conflict 
> between security and convenience:  The user tends to use fixed, easily 
> memorized keys again and again, while the cryptoanalyst only waits for an 
> opportunity to see many messages encrypted by the same key.  

I get the feeling you are confusing two separate issues here. Users do tend 
to use easily memorized passphrases, which is indeed a problem, because such
passphrases are in general easy to guess. But cryptosystems are generally
meant to resist attacks based on the cryptanalyst gaining access to many 
ciphertexts, even if all were encrypted with a single key.

[...]
> A simple example:  For a short message, increasing the font 
> size of the text by a factor of 10 will significantly increase the time 
> required for breaking the encryption.

Anyone know how to get 120 point text in LaTeX ?

> Customized versions of PrivaSoft
> 
> PrivaSoft is unique in being a one-stop product than can serve all types of 
> modern correspondence, including E-mail, fax and paper printouts.  Special 
> applications that need and can obtain a license to use non-commercial 
> cryptographic engines can be accommodated by special versions of PrivaSoft.  
> The cryptographic engine can be customer-furnished and customer integrated, 
> however - since in some areas the integration of this product with certain 
> cryptographic engines may be considered "munitions", each customized version 
> of the product has to be licensed separately in accordance with the laws of 
> the territory where it was created and used.

I think this is a smart idea.

-Futplex <futplex@pseudonym.com>