1. Mangai Vetrivelan Snigdha Joshi Avani Atre

2. Sensor Network Vulnerabilities o Unshielded Sensor Network Nodes vulnerable to be compromised. o Attacks on Sensor Networks:  Loss or Corruption of Sensitive Data  Disconnection of Networks

3. What is Node Replication? o Node Capture o Analysis (of Node Information) o Replication o Injection

4. Consequences of Node Clones o Tampering Data o Injection or Suppression of Node data o Revocation of Nodes o Disconnection and wrong configuration

5. How do we deal with Node Replication Traditional Approaches o Centralized Detection -Delayed Revocation -Base Station Compromise o Localized Detection -Failure for disjoint networks

6. Preliminary Approaches o Node-To-Networking  Each node broadcasts location to every other node  Node Revocation at each node on conflict Disadvantage: Extensive communication overhead

7. Preliminary Approaches Continued.. o Deterministic multicast  Number of nodes(n)  Each node broadcasts location to a subset of ‘g’ witnesses(nodes)  The subset of g nodes is for each node A is chosen by a function: F(A)={w1,w2,..,wg}

8. Deterministic Multicast Continued...  Each w checks for duplicate nodes(with different locations)  Replicated nodes are revoked

9. Deterministic Multicast Continued.. Disadvantages:  F is deterministic  Hence easy access to witness information  Adversary can determine and subvert witnesses Solution: Increase ‘g’ `Problem: Communication cost(c)=O(glngn/d) Thus, when g increases, C increases

10. Randomized Multicast  Every node sends its location to a subset of g randomly selected nodes  Improvement over Deterministic Multicast  Adversary cannot determine witnesses  Based on ‘Birthday Paradox’ The location claim of a node and its duplicate collides with a high probability. In a network of n nodes, root(n) witnesses are chosen. Collisions given by birthday paradox

11. Randomized multicast continued...  Each node sends the following message: Where Idα=Node ID of Node α lα=location of Node α Kα−1=private key of Node α  Each neighbour receives the message  Each neighbour forwards this location to ‘g’ random nodes with probability p

12. Randomized multicast continued  Each witness w{w1,w2,...wg} receives the message Verifies the digital signature against the ID Checks the ID against a list of Ids from the previous pre-stored location claims  If the ID matches any pre-stored ID Claim as duplicate Blacklist the node

13. Security Analysis of Randomized Multicast Consider a malicious node α, positioned at L locations l1,l2,l3…,ln We want to determine the probability of a collision using Randomized Multicast If neighbors are chosen in a coordinated way = alpha’s location claim is stored at p.d.g locations We want independent selection of nodes, so nodes might overlap. What is the impact of overlap ? No of witness nodes that have at least one collision. This is approximated to p.d.g.

14. Continued.. Not efficient, communication cost is O(n 2 ) P nc1 is the probability that the p.d.g recipients of claim l 1 do not receive any of the p.d.g copies of claim l 2 P nc is the probability of no collision at all. N = 10,000, g =100, d=20, p = 0.05, Probability to detect single replication is greater than 63%, Probability to detect two replication is greater than 95% If the adversary inserts L replicas of, we would like to determine the probability that two conflicting location reports

15. Line-Selected Multicast  When a node sends its location claim  Each intermediate node Stores the location claim Routes the claim  Exchange of location claims in such a manner corresponds to several intersecting communication lines

16. Line selected Multicast Continued  The node at the intersection: Compares two location claims with the same node ID Finds a mismatch Revokes using flooding

17. Sylvester’s Four- point Problem Pintersect=1/3(1- 35/12∏^2)=0.235(Approx)

18. Lets Compare Randomized Multicast Protocol & Line-Selected Multicast Protocol  Resiliency  Communication Overhead  Space Efficiency

19. Emergent Properties  Traditional algorithms: centralized  Emergent algorithms:  Collective efforts of multiple sensor networks  Highly distributed detection  Seen clearly in Randomized and Line-Selected Multicast Robust Resilient Low overhead

20. Conclusion  Randomized multicast and Line Selected Multicast perform efficient distributed detection of node replication.  Line-Selected Multicast is observed to be more efficient than Randomized Multicast in terms of: overhead, attack resistance  They use ‘Emergent Properties’

21. References http://research.microsoft.com/pubs/138312/replication.pdf

22. Thank You!