Very Computer

During the last days many people asked for an encrypted Telnet. After my
first posting some people were interested in getting the sources. Therefore
we put them on our FTP server. Here is the announcement:

You can fetch the package and some related stuff via anonymous FTP
from our server at the following location:

   ftp://ftp.tu-chemnitz.de/pub/Local/informatik/sec_tel_ftp

   (Internet address: 134.109.2.13)

Please read the file README first. This briefly explains what the
individual files in this directory contain and gives you some hints
how to install the package modified by us (tuc-if-sra.tgz).

Please note: the package srasrc-1.3.tar.gz was the starting point
             for our modifications. It is incomplete.
             It's not necessary to fetch it unless you are interested
             in the Kerberos authentication extensions we deliberately
             omitted or if some of our changes should be inconvenient
             for you.

===============================================================
Name       : Holger Trapp
Institution: Technical University of Chemnitz-Zwickau
             Faculty of Computer Science
             Chair of Computer Networks and Distributed Systems
Address    : 09107 Chemnitz
Location   : Strasse der Nationen 62
Phone      : +49 371 531 1379
Fax        : +49 371 531 1628

===============================================================

 [*] Holger Trapp writes:

   ] During the last days many people asked for an encrypted Telnet.

...and there are a couple of hacked versions of SRA announced. I
managed to add encryption to SRA during the weekend too. It turns out
to be a 10-liner.

Anyway, I propose we put our efforts into one definitive

the moment, because it's based on a newer version of telnet which
have been tested on a variety of platforms.

The other thing that we need to do is to change the root and modulus
of the exponential key exchange. Someone pointed out that the current
modulus is insecure. This however would mean that new versions of
SRA will not be compatible with older versions. We have to call it
SRA2 or something....

Comments?

Delman.
--
______________________________________________________________________

  Delman Lee                             Tel.: +1-215-662-6780
  Medical Image Processing Group         Fax.: +1-215-898-9145

______________________________________________________________________

Quote:

>Has anyone been able to get rsa telnet/ftp working WITHOUT NIS?

>I'm on a sun 4.1.3 system, and we don't have NIS (no secret keys :-()

>Any way around this?

--
Nick Christenson

Has anyone been able to get rsa telnet/ftp working WITHOUT NIS?

I'm on a sun 4.1.3 system, and we don't have NIS (no secret keys :-()

Any way around this?

   ]]
   ]] Has anyone been able to get rsa telnet/ftp working WITHOUT NIS?
   ]]
   ]] I'm on a sun 4.1.3 system, and we don't have NIS (no secret keys :-()

   ] Yes, the Encrypted Telnet I mention in an earlier post uses
   ] Diffie Hellman without the use of NIS or any RPC service by
   ] using the Public Key cryptography routines found in the RSAREF
   ] library.

SRA telnet/ftp does NOT require NIS or RPC service. It uses
Diffie-Hellman exponential key exchange to establish a DES key between
the client and the server. SRA uses a bit of Sun RPC source code to do
the Diffie-Hellman.

I haven't seen Nick's code, but I believe Nick's is similar to SRA's
telnet, differences being:

(i) Current implementation of SRA uses a 192bit modulus, which is too
small. ETN uses a >512 bit modulus. One could change the SRA code to
use a larger modulus and come up with SRA2.

(ii) ETN uses RSAREF code to do the Diffie-Hellman. SRA uses Sun RPC
code.

(iii) SRA telnet as distributed by TAMU doesn't do data stream
encryption. A 10-line patch will get that done.

(Any other differences, Nick?)

-----------------

My conclusions so far: SRA, ETN and code/spec mentioned by Jeff Hayward
all uses Diffie-Hellman to establish a common key. The main difference
being the modulus used --- SRA: small, ETN: medium/large,
Hayward:variable.

Since they all use Diffie-Hellman, I continue my plea for ONE
distribution that uses Diffie-Hellman. This would avoid duplication of
code and incompatibilities between different DH distributions.

About the patented Diffie-Hellman: would it be OK if US sites use
the RSAREF implementation, while non-US sites use another way like the
one in Sun RPC. Comments?

Delman.
--
______________________________________________________________________

  Delman Lee                             Tel.: +1-215-662-6780
  Medical Image Processing Group         Fax.: +1-215-898-9145

______________________________________________________________________

Quote:>I haven't seen Nick's code, but I believe Nick's is similar to SRA's
>telnet, differences being:

>(i) Current implementation of SRA uses a 192bit modulus, which is too
>small. ETN uses a >512 bit modulus. One could change the SRA code to
>use a larger modulus and come up with SRA2.

To change the SRA DH key size, I believe you would need to rebuild
the RPC library.  Again, please correct me if I'm wrong, somebody.

Quote:>(ii) ETN uses RSAREF code to do the Diffie-Hellman. SRA uses Sun RPC
>code.

Quote:>(iii) SRA telnet as distributed by TAMU doesn't do data stream
>encryption. A 10-line patch will get that done.

Quote:>(Any other differences, Nick?)

ETN negotiates the DES encryption keys (using DH) without recourse to
the standard telnet IAC will/wont do/dont mechanisms.

Quote:>-----------------

>My conclusions so far: SRA, ETN and code/spec mentioned by Jeff Hayward
>all uses Diffie-Hellman to establish a common key. The main difference
>being the modulus used --- SRA: small, ETN: medium/large,
>Hayward:variable.

>Since they all use Diffie-Hellman, I continue my plea for ONE
>distribution that uses Diffie-Hellman. This would avoid duplication of
>code and incompatibilities between different DH distributions.

Quote:>About the patented Diffie-Hellman: would it be OK if US sites use
>the RSAREF implementation, while non-US sites use another way like the
>one in Sun RPC. Comments?

Stupid, yes, but we have no choice.

--
Nick Christenson

<---- Header information -------------------------------------------->
Xref: actrix.gen.nz comp.security.unix:11891
Path: actrix.gen.nz!comp.vuw.ac.nz!waikato!ames!elroy.jpl.nasa.gov
!netline-fddi.jpl.nasa.gov!phoebe.jpl.nasa.gov!npc

 Laboratory)
Newsgroups: comp.security.unix
Subject: Re: ANNOUNCEMENT: Encrypted Telnet
Date: 20 Feb 1995 17:31:02 GMT
Organization: Jet Propulsion Laboratory, Pasadena, CA
Lines: 101



NNTP-Posting-Host: minotaur.jpl.nasa.gov
Cc:

Quote:>I haven't seen Nick's code, but I believe Nick's is similar to SRA's
>telnet, differences being:

>(i) Current implementation of SRA uses a 192bit modulus, which is too
>small. ETN uses a >512 bit modulus. One could change the SRA code to
>use a larger modulus and come up with SRA2.

To change the SRA DH key size, I believe you would need to rebuild
the RPC library.  Again, please correct me if I'm wrong, somebody.

Quote:>(ii) ETN uses RSAREF code to do the Diffie-Hellman. SRA uses Sun RPC
>code.

Quote:>(iii) SRA telnet as distributed by TAMU doesn't do data stream
>encryption. A 10-line patch will get that done.

Quote:>(Any other differences, Nick?)

ETN negotiates the DES encryption keys (using DH) without recourse to
the standard telnet IAC will/wont do/dont mechanisms.  

Quote:>-----------------

>My conclusions so far: SRA, ETN and code/spec mentioned by Jeff Hayward
>all uses Diffie-Hellman to establish a common key. The main difference
>being the modulus used --- SRA: small, ETN: medium/large,
>Hayward:variable.

>Since they all use Diffie-Hellman, I continue my plea for ONE
>distribution that uses Diffie-Hellman. This would avoid duplication of
>code and incompatibilities between different DH distributions.

Quote:>About the patented Diffie-Hellman: would it be OK if US sites use
>the RSAREF implementation, while non-US sites use another way like the
>one in Sun RPC. Comments?

Stupid, yes, but we have no choice.

--
Nick Christenson

   ] Based on the SRA paper from UNIX Security IV (I believe), it does
   ] not require RPC to be running, but does require that a library containing
   ] the Diffie-Hellman code be available on the machine SRA is compiled
   ] on.
   ] (...)
   ] To change the SRA DH key size, I believe you would need to rebuild
   ] the RPC library.  Again, please correct me if I'm wrong, somebody.

SRA compilation
===============

Compilation of SRA does not require the RPC library. SRA comes with
the source code to do Diffie-Hellman. It is taken from the source of
Sun RPCSRC 4.0 (legal issues at the end). SRA telnet should compile on
an HP and most systems cleanly as long as the underlying telnet code
compiles cleanly. Hope this clear things up about SRA compilation.

SRA modulus change
==================

The Sun RPCSRC uses the Multiple Precision library (SRA uses GNU's
gmp) to do the Diffie-Hellman. I think the code is quite general with
the modulus and stuff defined in a header file:

  #define HEXMODULUS "d4a0ba0250b6fd2ec626e7efd637df76c716e22d0944b88b"
  #define HEXKEYBYTES 48
  #define KEYSIZE 192
  #define KEYBYTES 24
  #define PROOT 3

Changing the above lines would result in a larger modulus. We just
have to decide on how large is large.

Legal issues
============

If I understand it correctly:

. non-US/Canadian sites can use the Sun RPCSRC source code.
. US/Canadian sites can use the RSAREF code or Sun RPC library.

Can US sites use the Sun RPCSRC source code? If not, then we will have
to modify SRA to use RSAREF for US sites.

Base on the above, I will start a new thread by posting a proposal for
a single distribution (based on SRA) that uses Diffie-Hellman. Please
follow up to that article.

Delman.
--
______________________________________________________________________

  Delman Lee                             Tel.: +1-215-662-6780
  Medical Image Processing Group         Fax.: +1-215-898-9145

______________________________________________________________________

Quote:>IMHO, too few systems have Secure RPC to make that the only "standard".
>For example, my HP-UX 9.0X boxes don't have it, and many of them
>will not be going to HP-UX 10.0X (which is supposed to have it)
>for many years.  It will not be a solution for me.  Does anyone know
>if Linux has Secure RPC?  I believe it is available on the free BSD
>systems.

Quote:>Non-US sites can use their own libraries without restriction.  They
>do not have the Public Key patent problems we do in the US.  In the
>US, we have to use either Secure RPC, RSAREF or pay for it.  Of course,
>it is illegal for folks in the US and Canada to share code that *calls*
>encryption routines with our friends in other countries, so US
>contributions to international standardization would have to be
>based on algorithm description rather than exchange of source code,

Alan

--
  ..-----------,,----------------------------,,----------------------------,,

 ``----------'`--[Anti Kibozing Signature]-'`----------------------------''
One two three: Kibo, Lawyer, Refugee :: Green card, Compaq come read me...

Quote:>(iii) SRA telnet as distributed by TAMU doesn't do data stream
>encryption. A 10-line patch will get that done.

   ]] (iii) SRA telnet as distributed by TAMU doesn't do data stream
   ]] encryption. A 10-line patch will get that done.

   ] Can anyone post this patch?

You could get my modified version of sra-1.3.1 at
ftp://ftp.mipg.upenn.edu/pub/delman/security/sra-1.3.1-modified.tar.gz

 telnet -x hostname

will give you an encrypted link.

Delman.
--
______________________________________________________________________

  Delman Lee                             Tel.: +1-215-662-6780
  Medical Image Processing Group         Fax.: +1-215-898-9145

______________________________________________________________________