North American Network Operators Group
Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: the O(N^2) problem

  • From: Edward B. DREGER
  • Date: Mon Apr 14 02:04:08 2008 
Stardate Mon, 14 Apr 2008, Suresh Ramasubramanian's log:
SR> From: Suresh Ramasubramanian

SR> Looks like what various people in the industry call a "reputation
SR> system"

I started responding; Suresh's reply came as I was doing so, and put it
very succinctly.  Reputation system, but inter-"network".  Perhaps an
example would work better than my vague descriptions. :-)

Let's say I receive email from:

	From: <[email protected]>
	Received: from ... (owen.delong.sj.ca.us [192.159.10.2])

Should I trust the message?  I don't "know" you.  However, I _do_ know

	From: <[email protected]>
	Received: from ... (trapdoor.merit.edu [198.108.1.26])

and trapdoor.merit.edu vouches for you.  Elaborating, using "trust
paths", *not* SMTP or routing paths:

	<[email protected]> # distrust metric: initially 0
	owen.delong.sj.ca.us # distrust metric: still 0
	trapdoor.merit.edu # dm: 1 (it mostly believes your MX)
	mail.everquick.net # dm: 2 (more or less trust NANOG)

versus

	<[email protected]> # dm: 0
	malicious.host.domain.tld # dm: 0 (trying to impersonate)
	trapdoor.merit.edu # dm: 10 (doesn't yet trust above host)
	mail.everquick.net # dm: 16 (after whatever local mods)

or

	<[email protected]> # dm: 0
	owen.delong.sj.ca.us # dm: 0 (your MX can vouch)
	trapdoor.merit.edu # dm: 1 (more or less believes your MX)
	mail.everquick.net # dm: 2 (more or less trust NANOG)

IOW, I receive email from an unrecognized address from your MX.  Do I
trust it?  I mostly trust trapdoor.merit.edu, which mostly trusts your
MX, which totally trusts <[email protected]>.  Therefore, I conclude
that I largely trust the message.

For such a system to scale, it would need to avoid OSPF-style
convergence.  Similarly, I would not want to query, for the sake of
example, 15k different "trust peers" each time I needed to validate a
new <host,address> tuple.  (Hence the interdomain routing and d-v calc
references.)

Therefore, one would find the locally-optimal solution at each "trust
hop", a la interdomain routing.

Perhaps PGP/GPG would be the best analogy.  (When I said "PKI", I should
have stated CA-based PKI; my original wording was excessively broad, and
should have explicitly excluded PGP.)


Eddy
--
Everquick Internet - http://www.everquick.net/
A division of Brotsman & Dreger, Inc. - http://www.brotsman.com/
Bandwidth, consulting, e-commerce, hosting, and network building
Phone: +1 785 865 5885 Lawrence and [inter]national
Phone: +1 316 794 8922 Wichita
________________________________________________________________________
DO NOT send mail to the following addresses:
[email protected] -*- [email protected] -*- [email protected]
Sending mail to spambait addresses is a great way to get blocked.
Ditto for broken OOO autoresponders and foolish AV software backscatter.