Denial of Service Resilience in Ad Hoc Networks

1
Denial of Service Resilience in Ad Hoc Networks

• Imad Aad, Jean-Pierre Hubaux, Edward W. Knightly
• (EPFL Rice)
• Appears in MobiCom 04
• Presented by David R. Choffnes

2
Outline

• Ranting
• More Ranting
• Zombies
• Ranting

3
JellyFish Attacks

• Requires a closed-loop protocol, like TCP
• But weve already discussed paper detailing how
  TCP doesnt work well for ad hoc networks
• How do protocol-compliant attacks work on
  cross-layer protocols?
• Not specific to ad-hoc networks
• E.g., periodic dropping
• It would be nice to compare performance against
  wired networks in these cases
• Can actually help the system when it is
  overloaded!

4
Methodology Attacks without Solutions

• Exploring impact of attacks without considering
  solutions lessens validity of results
• Authors do not consider more interesting
  scenarios such as hybrid environments (more than
  one TCP implementation)
• Even though nodes are protocol compliant,
  anomalous complete TCP collapse or significant
  reduction in performance is not hard to detect
• Seems to rely too much on cross-layer independence

5
Misbehavior Diagnosis

• PACK limitations
• Power management not likely in highly mobile
  scenarios
• Directional antennae not widely deployed yet, not
  necessarily true that they will be anytime
• Diagnosis time order of seconds
• No results back up the claim
• How does this change with different routing
  protocols?

6
Modeling and Simulation

• Mean route lifetime 10s
• Why not look at a range of mean lifetimes
  corresponding to real scenarios?
• Simulation environment
• Open field of 1.2 square miles!
• 200 nodes moving randomly! (Shaun of the Dead)
• Maximum velocity of 10 m/s (Most people cant run
  much faster than 3 m/s, esp. with a computer in
  hand, browsing the Web or using FTP)
• Receive range of 250m is ridiculously unrealistic

7
More issues with Experiments

• To what realistic scenario does the network
  traffic correspond?
• Figure 16 large radio range inhibits fairness
  measurement, but the authors do not consider
  reducing the radio range
• Only one routing protocol evaluated
• How does JF impact AODV?
• How about location-based routing protocols?
• Cross-layer reliable delivery protocols?

8
Finally

• This is NOT the Internet
• These are mobile scenarios, where all computers
  are relatively close to each other and in an open
  field
• Malicious attacker cannot simply hide behind a
  bush
• No doubt there are techniques to locate the JF
  attacker and physically remove her from the
  network