Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!uwvax!oddjob!gargoyle!ihnp4!ihnet!tjr
From: tjr@ihnet.ATT.COM (Tom Roberts)
Newsgroups: sci.crypt
Subject: Re: FIDONET Proposal - Public Key Encryption System
Message-ID: <562@ihnet.ATT.COM>
Date: Mon, 3-Aug-87 12:19:44 EDT
Article-I.D.: ihnet.562
Posted: Mon Aug  3 12:19:44 1987
Date-Received: Tue, 4-Aug-87 04:04:20 EDT
Organization: AT&T Bell Laboratories - Naperville, Illinois
Lines: 43
Keywords: Public Key Encryption

There are some problems with public key encryption which I do NOT
believe have been adequately addressed:

1) How does a user KNOW that the correct (i.e. uncompromised) public key
   is being used. This is crucial to the entire system. If anybody (e.g.
   SYSOP) can modify the keys contained in the key server, they can then read
   the mail, or masquerade as someone else. Of course, the recipient will 
   decrypt garbage, or the deception will eventually be discovered, but the
   damage has already been done.

2) How are keys selected? I know of no adequate method using psuedo-random
   number generators. Using the current date+time[+...] as a seed is hopeless.
   Using a "random" string entered by the user is better, but is still not
   perfect. Requiring the user to have a "true" random number generator is
   not feasible. Remember, the "seed space" for the key generator is the REAL
   "key space" which must be searched by a cracker; the 1000-bit secret
   key is merely an intermediate result determined by the key generator (but
   it, too, is vulnerable, so you need a good algorithim which makes the
   secret key uncomputable from the public key - see 3 below).

3) How secure is the algorithm (i.e. how "difficult" is it to determine the
   secret key, given the public key and the algorithm) ?
   There have been recent advances described in the unclassified literature
   which affect the popular algorithms.

Public key encryption (or any type of encryption) can make the reading of
mail by non-authorized people non-trivial. A general, system-wide 
implementation of any type of system is also non-trivial, even difficult.
Key management is where most systems fall down; contrary to statements made
in the "popular" press, public key techniques DO NOT solve the problems, but
merely move them to a different arena. The level of "non-trivialness" depends
heavily upon system design and implementation; strong algorithms do not
necessarily make strong systems - much more is involved.

Note the danger: users tend to believe that encryption systems are invulnerable
(c.f. all of the anecdotes about WW2, in which German and Japanese systems
were utterly broken, but they retained their blind belief they were secure).
It is entirely possible that a moderately-strong system will be WORSE than
no system, because it will engender a false sense of security in the users.


Tom Roberts
ihnp4!ihnet!tjr