From: Adam Back <aba@dcs.ex.ac.uk>
Date: Mon, 20 Jan 1997 02:45:26 -0800 (PST)
To: azur@netcom.com
Subject: Re: GSM crypto upgrade? (was Re: Newt's phone calls)
In-Reply-To: <v02140b00af086e8ed0db@[10.0.2.15]>
Message-ID: <199701170904.JAA00524@server.test.net>
MIME-Version: 1.0
Content-Type: text/plain



Steve Shear <azur@netcom.com> writes:
> For voice calls, it is not possible to merely insert a more secure crypto
> function (e.g., IDEA) inside the less secure A5 crypto, unless the GSM base
> station's protocol is aware the subscriber unit is using the more secure
> crypto.  Otherwise, when the GSM base station unwraps the A5 encrypted data
> stream, which it assumes will contain digitized voice packets in the clear,
> it will not find what its looking for and will be unable to convert the
> packets to a circuit-switched voice signal.

The circuit-switched voice signal is digital also, and the voice
packet payload is transferred byte for byte into the circuit switched
link right?  So to insert IDEA encryption inside the A5 layer, you'd
need to packetize the IDEA ciphertext so that it conforms to the voice
packet spec (to fool the unwrapping the GSM station performs into
thinking the packets are voice packets).  Is that possible?  Is the
packet switched protocol non-error corrected?  (ie Is packet dropping
expected?)  If so you need to use an IDEA mode which does not rely on
previous packets.

> If, as I have previously stated, the subscriber instead uses the data port
> of his instrument and establishes a data link with his payload protected by
> the more the secure crypto this is entirely feasible.

If the above is infeasible.  Probably this is a more practical
approach anyway, as the data port has been already engineered to solve
some of problems that would be encountered, and you would be
duplicating effort.

Adam
--
print pack"C*",split/\D+/,`echo "16iII*o\U@{$/=$z;[(pop,pop,unpack"H*",<>
)]}\EsMsKsN0[lN*1lK[d2%Sa2/d0<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<J]dsJxp"|dc`