[perpass] on encryption and social networks (Was Last call: draft-farrell-perpass-attack-02)

[Eliot Lear <[e-mail]>]

Hi Stefan,


On 12/11/13 8:35 AM, Stefan Winter wrote:
> Hi,
>
>>> Knives are easily available to anyone, just like encryption.
>> ...and just like pervasive monitoring?
> That's a very good thought.
>
> Yes, I believe encrpytion and the ability to pervasively monitor are
> both easily available to everyone.
>
> The next step after availability is actual usage, and this is where
> things get interesting.
>
> I believe that where encryption is not actually *used*, pervasice
> monitoring *will* happen. Or, to state it a bit more in a logic-oriented
> way:
>
> Either the use of encryption proliferates, or the use of pervasive
> monitoring proliferates.
>
> It is a strict XOR: you can't have both, and you can't have none of the two.

Pervasive monitoring is not just about encryption, and I suspect you
fully understand this, but have simplified your argument for the
purposes of focusing on encryption.  Even so, I like your analogy, in as
much as you are not looking at one big XOR but many little small ones
and then summing them on either side of the equation.  That's the nature
of engineering tradeoffs, to be fair.

In reviewing some other work on another list, I came across a paper that
talked about pervasive surveillance from an economics sense, written in
2006 entitled /The Economics of Mass Surveillance and the Questionable
Value of Anonymous Communications/.[1]  It's early work to be sure, and
actually argues that the adversary has to actually do very little to
achieve quite a lot in monitoring everyone.  But it also alludes to the
use of social network theory to answer questions of "who" not "what". 
While there are many different aspects of surveillance, social networks
are clearly a component.

Masking the "who" is much harder than the what, and yet the "who" may be
as or more valuable.  Unfortunately for us, the who can be represented
by IP address (as an example).  It's why there's some interest in TOR. 
Fortunately for us, use of services like Google or Facebook or Twitter
provide a means to obscure paths of communication.  Unfortunately for
us, they become single points of failure.[*]

This goes to a point that Mark Nottingham made in the plenary.  How do
we strike a balance between those single points of failure on the one
hand and the potential benefits that some amount of aggregation can
bring?  There has been other related work on this topic from the context
of platform diversity and cybersecurity, but I regret that I couldn't
find the paper I wanted to cite for this email (I tried- it's there-
maybe others can dig it up) that is worth some consideration.

My point in raising this now and my point of intervening several times
in the plenary is to highlight that many of the issues around pervasive
surveillance are complex and have many tradeoffs to be made.  There's
some research to be reviewed, and perhaps some research to be
performed.  When we say that we the IETF are going to do something, we
need to be looking at a long term view, while taking reasonable actions
in the short term that heads us in the right direction.

Eliot

[1] http://weis2006.econinfosec.org/docs/36.pdf
[*] I imagine there's an undiscovered Rogers and Hammerstein song called
"Fortunately/Unfortunately" buried in here.