21 March 1997

To: cypherpunks@toad.com
From: lucifer@dhp.com (lucifer Anonymous Remailer)
Date: Thu, 20 Mar 1997 06:02:43 -0500
Subject: Naval Intelligence/Traffic Analysis/PGP Security

I am passing along the following message, not because of my belief in the accuracy or veracity of its contents, but merely because of the fact that it seems to have kept getting eaten up by various email systems in the attempts to send it to its original destination.

The first time it was sent via a remailer, it was bounced for ill-defined reasons. The second time it was sent, the remailer was shut down, and remains shut down.

Efforts to send it through a second remailer also failed, with no notice from the server of any problems being received. Other email sent through the remailer at the same time encountered no difficulties.

I am only noting the facts, here, not proferring any personal judgments on the matter, although I have formed my own opinion as to the meaning of these facts, particularly since I, myself, came by this post during the process of email interception.

----------Forwarded Message----------

> X-Anon-Password: XXXXXXXXXX


I thought I would reply privately to you, since you seem to at least have a willingness to allow the possibility of compromises to the security of the encryption methodologies behind PGP programs, among others.

To begin with, I'm not sure whether you realize it, or not, but the Navy's spook tenacles run deeper, and extend further, than those of any of the more notable or visibly involved agencies who lurk in the background of security and privacy issues.

One of the reasons for this is that their physical existence could be said to mirror the Internet in many respects. The very nature of their 'global' home (the sea), has always permitted them access to people and regions which are denied to others. Also, they are often in the position to be involved in what looks to be merely the 'transporting' of people and information.

Whether providing escort services or getting drunk in foreign bars, the expertise of naval intelligence has always lain in the area of observation, first and foremost.

By far the greatest tool of intelligence agencies on the Internet, has been traffic analysis. Their techniques are sufficiently sophisticated that I would not be surprised to find out that they can tell more about us from our Internet activity than can be learned from the satellites capable of reading the newspaper over our shoulder as we sit in the park.

Traffic analysis involves all measurable quantum of information, the chief concerns being the patterns and timing of data transfer, from which everything ranging from content and motivation can be deduced.

If you wish to think in terms of back-doors, then you would be well advised to go beyond the concepts of 'passwords' and 'holes' and try to think in terms of patterns and timing, and other such 'structures' which are peripheral to concerns regarding 'code' and 'mathematics.' i.e., As well as considering the 'content' of what a program returned, you must also consider 'when' the program returned the result, and the patterns in the timing, as well as the content.

An analogy could be made to a person who, being interrogated, answers all questions with a predictable rhythm and then 'pauses,' however slightly, in answering a certain question. You can see that what is revealed by the 'content' of the answer can be greatly insignificant compared to what is revealed by the 'delay' in answering.

To expand your concept of 'back-doors' and 'holes,' you have to ask questions such as:

"Does it take a program or hardware longer to return a result of '0', than to return a result of '1'?"

"What factors can be introduced into the hardware and/or software that can influence the patterns and/or timing of various processes and the results they return?"

"Can key searches be made more efficient by analyzing such things as rhythm, syntax, etc? What 'details' or 'qualities' of an individual, group, or 'arena of concern' can be analyzed for the purpose of being able to group them into structures which can be searched for?"

"How can 'assigning' a value to certain sequences of numbers be used as a pattern to 'filter' the input data into a form which is easier to analyze?"

You are aware of 'tricks and techniques' that apply to mathematics and are widely known. i.e., The process of shifting and adding numbers when multiplying by the number '11'.

However, what about those quantum of information which are of no consequence to those seeking for the 'final result' of that multiplication? Can the peripheral effects of mathematic calculations be used to analyze what has taken place, to narrow the scope of inquiry?

My nephew describes numbers as getting 'wider' as they get larger, and he does quick checks of his result through his 'feel' for how much 'wider' a number should be when he is done, even in complicated equations which he ill-understands.

(He reminds me of Steven Wright, who claims that someone told him that his socks didn't match, and he replied, "Sure, they do. I go by thickness.")

I am currently working on a project which involves merging chaos theory with traffic analysis and other processes to analyze the effects that algorithms display when processed through the filters of varying hardware and software structures and methodologies.

The RSA algorithm and accompanying RSAREF subroutines were our first focus, for the very reason that there were certain factions behind the scenes of the Zimmerman/RSA agreement who seemed to have an inordinate amount of interest in the subroutines being chained to the algorithm (for reasons that have nothing to do with patent protection).

Those whose expertise goes far beyond my own in this area look at the initial results of the analysis as confirming that their is a 'relationship' between the RSA algorithm and the RSAREF subroutines which will enable them to break the system down into workable units for fairly quick analysis.

What is interesting is that the results from small probes into other encryption systems show the same potential for exploitation using varying analysis methodologies and processes. (One fairly well-known encryption routine is almost lame enough to reveal its secrets to anyone with a pencil and a stopwatch, as well as the file size and time it takes to encrypt.)

While I would rather you didn't publicize the preceding information, as a general rule, I think that is something that should be shared with anyone who is seriously focusing their efforts on better methods of encryption and analysis of encryption methodologies.

I am aware of two other groups who are working along the same lines, although with a narrower range of variables than ourselves, and I am certain that there must be more than a few other entities out there who are also pursuing this line of research.

I would appreciate any comments you may have on the above, as well as any suggestions you may have. (Despite having a post-graduate degree in an area which required a thorough grounding in mathematics I may have to refer any highly technical suggestions to those in the group who dream in numbers, sunset to sunrise.)

----------End Forwarded Message----------

[End anonymous message]

Note: The National Research Council's report, Cryptography's Role In Securing the Information Society, recommends that national security and law enforcement agencies enhance technologies and training for traffic analysis, cryptanalysis and other unidentified systems in lieu of asking for continued governmental restrictions on robust encryption. And, the Defense Science Board's report, Information Warfare - Defense, claims that encryption should not be considered a primary concern.