6 October 2002

Date: Thu, 3 Oct 2002 21:19:22 -0400 (EDT)
From: Steve Bellovin <smb@research.att.com>
To: cryptography@wasabisystems.com
Subject: Confidentiality as a goal of old telegraph codes

It's a truism in the crypto business that the old telegraph codes were for economy, with confidentiality against casual readers a noted and desirable goal.  But I've recently acquired two old codebooks that have stronger ambitions.

The more interesting one is Slater's Telegraph Code, since confidentiality is its only goal.  I have the 9th Edition, from 1938, but it appears to be originally from the late 1860's.  It encodes 25,000 words, including "a" and "the".  There are no sentences, phrases, etc.  Users are told to convert the plaintext word to a number, transform the number, and convert back to a new word for transmission.  Suggested transformations include adding or subtracting a shared secret constant, permuting some of the digits of the code number, and/or regrouping the digits of a string of code numbers.  Clearly not military-grade security, even for the time, I'd guess; in addition to the rather simple transforms, it's a one-part code.

Equally interesting is the threat model.  I quote from the introduction:

On the 1st February, 1870, the telegraph system throughout the United Kingdom passes into the hands of the Government, who will work the lines by Post Office officials.  In other words, those who have hitherto so judiciously and satisfactorily managed the delivery of our sealed letters will in future be entrusted also with the transmission and delivery of our open letters in the shape of telegraphic communications, which will thus be exposed not only to the gaze of public officials, but from the necessity of the case must be read by them.  Now in large or small communities (particularly perhaps in the latter) there are always to be found prying spirits, curious as to the affairs of their neighbours, which they think they can manage so much better than the parties chiefly interested, and proverbially inclined to gossip.

It goes on to warn of the need for confidentiality in business communications, especially when undersea telegraph lines are used.

Equally interesting is the fact that despite the common wisdom that says that secrecy products didn't sell well, this book survived for about 70 years -- with my edition being printed on the eve of war.

The other confidentiality code I have is "Sheahan's Telegraphic Cipher Code", from 1892.  It was intended for use by railway labor organizers, to keep management from knowing what they were up to.  It has about 7000 code words.

It's a more conventional telegraph code, in that it includes some phrases.  The general confidentiality scheme is similar to Slater's, though the only suggested transformation is adding or subtracting a constant to the code number.  Because the plaintext is phrases, rather than just words, there are separate code words along with the code numbers; these words are sent, rather than the numeric values.

From a cryptographic perspective, the most interesting item is that times, days, and numbers do not have code numbers -- the instructions say to send just the code words.  The compiler was worried about a known or probable plaintext attack on the offset value used for superencipherment.  There is also a warning against mixing plaintext with ciphertext, "excepting the name of a person or the name of a town". There is a cipher alphabet for spelling out words, but it, too,  is not superenciphered.

Some of my other, larger code books could have been used in a similar fashion, but there's no hint of that in the instructions.

--Steve Bellovin, http://www.research.att.com/~smb (me)

http://www.wilyhacker.com ("Firewalls" book)


The Cryptography Mailing List

Unsubscribe by sending "unsubscribe cryptography" to majordomo@wasabisystems.com