Slide 1: traffic observation in a stateless

traffic observation in a stateless 
data networking environment


CRISP/CISAC's IC3T
7 dec 99

kc, UCSD/SDSC/CAIDA
kc@caida.org 
www.caida.org

Slide 2: outline

outline


define stateless

what to protect, how to parameterize
what is the threat model

what can we realistically measure
what technical facilities do we have

need to protect user privacy
need for community buyin on an international scale
need for secrecy

Slide 3: stateless

stateless

cannot assume any end-to-end state 
         (unlike telephony systems)

little (no) facility for realistic tracking 
high probability (assured) attribution distortion

IP 
self-contained 
"fire and forget" 
no acknowledgments required -- source forgeable

          .... all By Design

Slide 4: what to protect?

what to protect?


how to parameterize the problem
 
end systems
routing infrastructure
non-networking/non-hosts objectives
       (funky political stuff i don't get)

Slide 5: what is the threat model?

what is the threat model?


who are we protecting against

bored (albeit brighter/quicker than we) kids and high school hackers
criminals with Teleological Objectives
good guys, probably overworked
       (have you tried to configure a router lately?)

  relative threat of malice vs incompetence
   (damage * intention) product way higher for latter

   --> system just plain fragile

Slide 6: what can we realistically measure?

what can we realistically measure?


1) active tagging 

2) passive observation

both require

vendor & community buyin (+laws)
phenomenally complex
       data collection infrastructure

Slide 7: what technical facilities have we?

what technical facilities have we?


routing/switching equipment 
can be modified if needed (at cost)

performance measurement infrastructures 
active monitoring
many specialized ones
much weak, inconsistent methodology
workload characterization infrastructures
NLANR (NSF-sponsored HPC sites)
passive monitoring

     ... none exist enuf for 
          global traffic tracking

Slide 8: CoralReef

CoralReef 


most appropriate available platform 
high flexibility wrt dynamic requirements
public
US tax dollars help fund 

hard part remains
define (realistic) requirements
deploy infrastructure
synchronize collections 
    (through multiple measurement points)

Slide 9: need to protect user privacy

need to protect user privacy


IP header fields (esp. IP addresses)
IP data content

intertwined w encryption
tremendous opportunities for abuse, 
           if/as implemented
govts must defend need for capability
     to screw over every citizen,
  provide assurance that "terrorism capes" 
    won't be used against them

...or they'll just `route around the damage'

Slide 10: need for secrecy?

need for secrecy?


communicating attack profiles

attack by outlaws on others
attack by law enforcement on non-outlaw citizens

should terrorism counter-measure architecture/implementation be public?

secrecy complicates agenda

Slide 11: need for ISP community support

need for ISP community support

o/w just non-starter at best
insulting at worst
ISPs work way harder than we do
most are (after capitalist) libertarian
many disagree with us
or just think you're reality-detached
seriously, talk to some of them

   ``...insanity in the face of a bunch of war nuts and cops 
   trying to build a regime that can regulate what 
   it's taken us 30 yrs to deregulate 
   (end-to-end free expression independent of media or borders)''

   ``....  governments have killed far more people than 'terrorists' 
   or citizens ...  public acceptance of the Internet may perhaps 
   mean that we who built it have somewhat of a clue as to 
   how the world should be governed (or not governed, as the case may be)''

   ``danger of govt by the clueless, over a place they've 
   never been, using means they don't possess''

Slide 12: www.caida.org/Presentations/

www.caida.org/Presentations/




k claffy 
UCSD/SDSC/CAIDA
kc@caida.org
www.caida.org