prev

next

of 24

View

37Download

3

Embed Size (px)

DESCRIPTION

Securing Distributed Sensor Networks. Udayan Kumar Subhajit Sengupta Sharad Sonapeer. Flow. Obstacles Security requirements Attacks Defense A probabilistic approach towards key management Base Station Security. Obstacles. Very limited resources Memory, power Unreliable communication - PowerPoint PPT Presentation

Securing Distributed Sensor NetworksUdayan KumarSubhajit SenguptaSharad Sonapeer

FlowObstaclesSecurity requirementsAttacksDefenseA probabilistic approach towards key managementBase Station Security

ObstaclesVery limited resourcesMemory, powerUnreliable communicationUnreliable transferConflicts while broadcastsLatencyUnattended operationPhysical attacksManaged remotelyNo central point management

security requirementsData confidentialityData integrityData freshnessAvailabilitySelf OrganizationAuthentication

AttacksSybil AttackTraffic analysis attackNode Replication attackAttack against privacy

DefenseFocus on two methodsKey managementProvides for data confidentiality, integrity, freshness and authenticationSecuring base stationTraffic analysis attacks

A probabilistic approach towards key management DSN Nodes have limited computation and communication capabilities.

DSN a truly dynamic infrastructure.

So traditional approach is vulnerable and impractical.

FACT: Energy consumption for a RSA (1024-bit) is about 42 mJ whereas for a AES it is 0.104 mJ in Motorola MC68328 (a mid range processor).

Solution Approach DSN node is given a key-ring of size k randomly chosen from a key pool of size P before deployment.

Because of the randomness; two sets of k keys may be completely different.

If a path of nodes sharing keys pair-wise exists then that path is used to exchange key, thus establishing a direct link.

Key Pre-DistributionA large pool of P keys (~ 220) and their identifiers are generated.

k keys are drawn randomly without replacement to construct a particular key-ring and loaded to a node of DSN.

A trusted controller node saves the key identifiers of a key ring and associated sensor identifier.

only a small number of keys needed to ensure that any two nodes (at least) share a key with a certain probability.

Experimental result shows that, for a probability = 0.5, only 75 keys drawn randomly out of a pool of 10,000 keys need to be on any key ring of a node.

Shared-key discoveryGoal - discover the node with which it shares a key.

The easiest way - Broadcasting.

Hide key-sharing patterns among nodes from an attacker and establish private shared-key discovery.

The recipient decrypts it with the proper key.

Creates the routing topology that guarantees the existed secured link, as a link implies sharing of a key. Also sharing of 2 or more keys between sensor nodes doesnt cause a link security exposure.

Path-key EstablishmentA path-key is assigned to selected pairs of sensor nodes that do not share a key.

But they are connected by two or more links at the end of the discovery phase of the shared-key.

key-ring size (k) is determined anticipating the fact of revocation and incremental addition of new sensor nodes, since both may require the execution of the path key establishment phase after shared-key discovery.

Some issues of DSN

Revocation.

Re-Keying.

Resiliency to node capture.

Analysis

p = prob. of existence a shared key between 2 nodes.n = number of nodes.d = p*(n-1) = expected number of edges connecting that node with its neighbor.

Now we will try to find d so that DSN will be connected.

We also want to determine the pool size of keys (P) given a limit for k keys in each node for a DSN of n nodes where d is given under a neighborhood connectivity constraint (say n = neighborhood connectivity of a node n

Analysis(contd.)Pc = lim prob. [G (n,p) is connected] = exp (exp(-c)) n-> infwhere p = (ln(n) /n) + (c/n) [c is any Real constant]

p = d/( n - 1) >> p. So p precisely gives us the probability that 2 nodes share at least a key from their k sized key-ring that was chosen from a pool of size P [not a sensor design constraint and may be very big]. Given n we can find p so that G is connected with Pc .We have to find out P for a given k and for a p .

Analysis(contd.)p =1-prob. ( two nodes dont share a key) = 1 - (P-k ) C k / P C k

Using Sterling Approximation : n ! (2)1/2 (n)n+(1/2) e-n

So we have, p = 1- [(1-k/p)2(P-k+(1/2)) / (1-2k/P) (P-2k+(1/2))]

Important ConclusionsSize of a DSN (n) has little effect on the expected degree of a node required to have a connected graph.

If P = 10,000 then only k = 75 keys are required to be distributed to any two nodes to make p = 0.5 to share a key from their key ring. Now for k = 250 if we take P = 100,000. This proves the scalability.

Almost certain connectivity through shared-key for a 10,000-node DSN, a key ring of size only 250 have to be pre-distributed.

Base Station Security

Multi-path routing to multiple base stations

Confusion of address fields

Relocation of base station

Multiple Base StationsRoute Discovery

Route Request

Route Feedback

Multiple Base StationsMulti-path data routing

Compute the connectivity information from the feedback messages

Compute global topology of the network

Compute redundant routes for each node

Construct forwarding tables for each node (forwarding table entry for each route node lies)

Dispatch the forwarding tables

Multiple Base StationsMulti-path data routing (contd) (Computing 2-redundant routes)

Choose two independent paths for any desired node A

First path to the closest base station (Use BFS)

Second path to any base station (Three s1, s2, s3 sets of nodes)

Disguising Base station locationDuring route discovery

Reversible hash function H(x) , shared key Kc

For each ID m, computeCm = {x: H(x) = m}After route discovery

Pair-wise keys for each neighbor nodes on the same routeSent along with the forwarding tables

Base Station RelocationUniform Random Deployment

Attack on vicinity of Base station

Both Base stations on the opposite edges

Base Station RelocationDensesparse Graph

Attack on the center of the dense part

One Base station on dense-sparse edge

Other Base station on opposite to first

Thank You

From traditional security. Why cant we use traditional security systems.