[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Who "invented" remailers?




At 1:06 AM -0700 6/4/97, Lucky Green wrote:
>At 06:27 PM 6/3/97 -0700, Tim May wrote:
>>(There are important issues, discussed by several of us several years ago,
>>and more recently by Wei Dai and Lucky Green, dealing with correlation
>>analysis of messages sent and messages received...esentially pattern
>>analysis.
>
>This type of analysis is unlikely to provide useful results for
>high-latency systems such as remailers. However, this analysis can be very
>fruitful on near real time systems such as ill-designed chained http
>anonymizers.
>
>[Tim of course knows this, I just want to make sure that this is
>universally understood.]

Agreed, but there remains a risk even for conventional remailer uses. It
all depends on latencies, numbers of messages, etc.

Let us imagine Lucky and I are corresponding through a remailer network.
Imagine the average latency, in total, is 2 hours. (Not unreasonable.)

The kind of pattern analysis I'm talking about would look for "digrams"
between my sending a message and Lucky receiving one.

--Tim sends a message, a couple of hours later Lucky gets one.

--then nothing, and Lucky gets nothing for several hours or more

--Lucky sends a message and a couple of hours later Tim gets one

--then nothing for several hours or more

--Tim sends one, Lucky gets one a couple of hours later

It all depends on:

- how many other messages are being received by Lucky and Tim (PGP messages
arriving from remailers, obviously, not just ordinary traffic, though
ordinary traffic helps a bit)

- the latencies, the longer the better (related to the above number)

What I am picturing is a scatter plot of transmissions and receptions. I
think an adversary with access to the sends and receives, even if encrypted
(of course) could make some plausible deductions.

He could certainly rule out some message pairs, e.g., Tim sends a message,
but no message is received by, say, Perry. Or Perry sends a message, but
this is not followed within statistically expected periods by messages
received by, say, Duncan.

So the "Tim-Lucky" digram might be "0. 6" and the "Tim-Perry" digram might
be "0.003" and this would be useful in implicating likely comunicants.
Throw in the reverse paths, e.g., the "Lucky-Tim" digram and the
correlations could become quite strong.

(Closely related of course to traffic analysis in general.)

As we have been saying for years, there's a nice MS thesis in this for someone.

--Tim May

There's something wrong when I'm a felon under an increasing number of laws.
Only one response to the key grabbers is warranted: "Death to Tyrants!"
---------:---------:---------:---------:---------:---------:---------:----
Timothy C. May              | Crypto Anarchy: encryption, digital money,
[email protected]  408-728-0152 | anonymous networks, digital pseudonyms, zero
W.A.S.T.E.: Corralitos, CA  | knowledge, reputations, information markets,
Higher Power: 2^1398269     | black markets, collapse of governments.
"National borders aren't even speed bumps on the information superhighway."