1. Toward Resilient Security in Wireless Sensor Networks Rob Polak Feb 23 2006 CSE 535

2. What is Wireless Security on the Link Level?  Message Authenticity  Verify Sender  Verify Message has not been forged  Message Privacy  The messages can not be read by a third party.

3. Previous research  Pairwise Key Distribution  Nodes contain a pool of symmetric keys, with a probability they contain shared keys.  These shared keys are then used to create a pairwise key used to endorse messages.  What are the problems with this method?

4. Problems with Pairwise  As more nodes are compromised the fraction of affected pairwise keys increases quickly.  Insider Attacks are not accounted for in the system.  Some sensors may not be able to communicate if they do not share keys.

5. Solution?  Location-Based Resilient Security (LBRS)  Split terrain into grids, and use a locally binded key

6. Overview LBRS  When an event occurs it is endorsed by multiple nodes within a cell.  Message is then forwarded to a node up stream towards the Sink.  Messages are verified en-route to ensure validity.

7. Grid Construction  How to construct a grid with no real infrastructure.  Solution: construct a virtual grid of cells, and bind keys to certain cells.  How to determine cell size? What are the tradeoff’s?  As cell size increases nodes are required to have less keys, however, if a large cell is compromised an attacker can forge events of a larger area.

8. Bootstrapping  Time when node is first deployed, and needs to generate it’s keys  Node determines its position  Node generates keys based upon its location, a master secret, and a one way function.  Then the node identifies all of the nodes in its sensing range and generates keys for those nodes. (used later in en-route message filtering)  Master secret is then erased permanently (no more keys can be generated).

9. En-Route Filtering  Any given report requires (m-1) distinct MAC endorsements (message authentication codes)  Reports are collectively processed and endorsed by surrounding nodes within a cell.  Once a message is sent to it’s upstream node (using geographic routing) the senders mac’s is then verified by the receiving node.

10. Routing  LBRS uses a concept of beam width routing, which is a subset of a geographic routing.

11. Analysis  Analysis Info  Given: a circular terrain of radius R and N sensor nodes  For fabricated attacks where m-1 distinct MAC’s are needed to verify a report the detection ratio is : 1 - ½^(8s(m-1)) = 0.999 =99% detection rate for our simulation.  In a simulation network of 10km with 400K nodes, the forged reports were found in an average of 4.2 hops, and 6 hops at most.

12. Node Compromise  Can we prove our hypothesis that LBRS is less vulnerable to node compromise.  Results from the simulation show that when 100 nodes are compromised only 11 cells or 0.68% of the total terrain. (30k nodes)  No comparisons to pairwise system.

13. Implementation  Implementation  Only talks about very basic setup of nodes.  Seems to be “missing” any results.

14. Future Work  Implementing the system and study the performance

15. Discussion  What are some of the problems with this system?  Can not handle networks with nodes that change location.  Does not scale well into system with low density of nodes.  Is this a viable network security solution? Are you convinced?