Two Tier Routing Protocol for Heterogeneous Sensor Networks

1. Two Tier Routing Protocol for Heterogeneous Sensor Networks Presented By: MdTariqul Islam Pavel Computer Science University of Kentucky

2. Most Existing Works Consider Homogeneous Sensor Networks • All sensor nodes are modeled to have the same capabilities in communications, computation, memory storage, energy supply, reliability and other aspects. • With pure static clustering, it is evident that the cluster head nodes will be over-loaded with the long range transmissions to the remote base station. The cluster head nodes expire before other nodes • Role-rotation can be used to solve the problem, but possess the necessity of increasing the hardware capabilities for each nodes

3. Routing Protocols & Security Issues • Most existing routing protocols considered routing protocols & security issues separately • Few of them considered security issues during the design time of a routing protocol • It’s nontrivial to fix the problem that routing protocol can be made secure by incorporating security mechanisms after the design has completed

4. Heterogeneous Sensor Networks (HSN) Model

5. HSN Model HSN model consists of 2 different types of nodes: Low-end sensors (L-sensors) Large in number Not equipped with tamper-resistant hardware Low transmission range High-end sensors (H-sensors) Small in number Equipped with tamper-resistant hardware Powerful in computation and energy Longer transmission range Higher data rate Base Station (BS) Well protected and trustable

6. Cluster Formation In HSN • Both L-sensors and H-sensors are uniformly and randomly distributed in the field • After sensor deployment, clusters are formed in an HSN • Powerful H-sensors serve as cluster heads and form clusters around them • All the H-sensors form a backbone in the network • Both H-sensor and L-sensor know their location information

7. Two Tier Secure Routing Protocol • The BS, H-sensors and L-sensors form a hierarchical network architecture • Two-Tier Secure Routing (TTSR) protocol architecture consists of two parts: • Secure routing within a cluster(among L-sensors)  Intra-cluster Routing • Secure routing across clusters(among H-sensors)  Inter-cluster Routing

8. Secure Intra-Cluster Routing • Routing from an L-sensor to its cluster head (i.e. H-sensor) is referred as Intra-Cluster Routing • Each L-sensor has one shared-key with every neighbor L-sensor • Consider two neighbor L-sensors u and v, and denote their shared-key as Ks. Assume node ID u<v. L-sensors u and v need to perform the following two-way handshake before exchanging any data:

9. Secure Intra-Cluster Routing • N0 is a one-time random number generated by u • MAC(Ks ,∗) denotes the Message Authentication Code (MAC) generated from message using key Ks. • Ku,vand Kvb are keys generated by v • Ku,vis the new pairwise shared-key used for the later communication between u and v • Kvbis a broadcast key for v u v

10. Secure Intra-Cluster Routing This two-way handshake can avoid (or defend against) the unidirectional link problem (or attack). For example, if u is a more powerful node (such as a laptop with a longer transmission range) than v, then u can send a packet to v directly, but v can not send a packet to u in one-hop. However, node v still thinks that u is a one-hop neighbor, and various problems may arise. u v

11. Secure Intra-Cluster Routing: Routing Tree Formation Let all L-sensors in a cluster form a tree rooted at the cluster Head • To minimize the energy consumption: • Completedata fusion→MST • i.e., two k-bit packets come in, and one k-bit packet goes out after data fusion. • Nodata fusion within the cluster→SPT • Partial data fusion→ NP-Complete problem • If data from nearby sensors are highly correlated, then an MST can be adopted to approximate the least energy consumption case. • Centralized algorithm can be used to construct an MST, so does SPT.

12. Secure Intra-Cluster Routing: Routing Tree Formation & Data Forwarding • During route setup, each L-sensor may record two or more parent nodes. Each L-sensor also records one or more backup cluster heads since these nodes are prone to failure • Following secure data forwarding scheme is used by L-sensor: • u → v: packet_ID + {Data}Ku,v + MAC(Ku,v, ∗) • L-sensor is responsible to guarantee the delivery • u will re-transmit if it does not get an ACK • The process continues until the packet reaches the cluster head H

13. Secure Inter-Cluster Routing • After cluster formation, each cluster head exchanges location information with neighbor cluster heads • During route discovery, H-sensor draws a straight line L between itself and the BS • The double dotted line intersects with a series of the clusters C0, C1, ...,Ck, which are referred to as Relay Cells • Packets are forwarded from the source cluster head to the BS via cluster heads in the Relay Cells using a secure data forwarding scheme similar to Intra-cluster Routing

14. Secure Inter-Cluster Routing: Reliability • H-sensors are more reliable nodes than L-sensors • If any cluster head in the Relay Cells is unavailable, then a backup path is used • Use self-healing scheme for H-sensor failures • Use a detoured path to avoid the failed cell

15. Two Tier Secure Routing Protocol:Security Analysis • Security Configuration • Data Authentication and Data Integrity • Achieved by MAC • Data Confidentiality • Achieved by symmetric encryption

16. Security Analysis: Defending Against Attacks • TTSR can defend against various attacks on sensor network routing • Sybil Attack • Wormhole Attack • Sink-hole Attack • Selective Forwarding Attack

17. Sybil Attack • In Sybil attack, a single node presents multiple identities to other nodes in the network • Authentication with pairwise shared-key is used in TTSR to ensure one node cannot pretend to be other nodes

18. Wormhole Attack • An attacker has two trusted nodes in two different locations of a network with a direct link between the two nodes • Replaying valid network messages at improper places, wormhole attackers can make far apart nodes believe they are immediate neighbors, and force all communications between affected nodes to go through them

19. Sinkhole Attack • The adversary’s goal is to lure nearly all the traffic from a particular area through a compromised node • The compromised node advertise an extremely high quality route to a base station

20. Defending Against Wormhole & Sinkhole Attacks • In TTSR Intra-cluster Routing, L-sensors send data only to its parent node of the (MST or SPT) tree and parent-child relationships are set up by the cluster heads (H-sensors) only • For TTSR Inter-cluster Routing, packet is forwarded only by the H-sensors in the Relay Cells. Therefore TTSR is resilient to both Wormhole and Sinkhole Attacks.

21. Selecting Forwarding Attack • A malicious node selectively drops sensitive packets • H-sensors are protected by tamper-resistant hardware. So, H-sensors can not be compromised • L-sensors might be attacked, but packet_IDfield is used to defend this attack

22. Evaluation of Routing Performance

23. A. Routing Performance under Different Node Densities Delivery Ratio (%) • TTSR has higher delivery ratio than DD (In TTSR, an L-sensor only needs to send packets to its nearby cluster head, and the rest transmissions are done by the H-sensor backbone.)

24. A. Routing Performance under Different Node Densities Energy Consumption • Energy consumption of DD increases much faster than TTSR, and it becomes very large when node density is high • (in DD more and more nodes are involved in disseminating “interest“ and “gradient" when node density increases)

25. A. Routing Performance under Different Node Densities Delay (ms) • TTSR has smaller end-to-end delay than DD for all the tested sensor density (The reason is that the same pair of source-destination in TTSR uses fewer hops of transmissions than that in DD)

26. B. Routing Performance for Different Source-BS Distances

27. B. Routing Performance for Different Source-BS Distances Delivery Ratio (%) • For any source-BS distance, the delivery ratio of TTSR is higher than DD (TTSR utilizes H-sensors for most transmissions and thus has less hop count than DD)

28. B. Routing Performance for Different Source-BS Distances Energy Consumption • Energy consumption of DD increases much faster than TTSR, and it becomes very large when the distance is large (In DD, more nodes participate in routing as the source-base station distance increases, In TTSR, only the number of L-sensors involved in the intra-cluster routing increases, while the number of H-sensors for the inter-cluster routing remains the same)

29. C. Routing Performance for Different Node Failure Probabilities

30. C. Routing Performance for Different Node Failure Probabilities Delivery Ratio (%) • The decrease in delivery ratios for TTSR is much slower than DD for different L-sensor failure probability (For the same source-BS pair, fewer sensors are in the route in TTSR that those in DD. In addition, H-sensors are less likely to fail)

31. C. Routing Performance for Different Node Failure Probabilities Energy Consumption • Energy consumption of DD decreases as node failure probability increases, since fewer sensors are involved in routing as more nodes fail. The energy consumption of TTSR increases a little bit as probability increases; due to node failures that cause re-transmissions and additional security operations in TTSR.

32. Performance Summary TTSR has: • A higher delivery ratio • A smaller end-to-end delay • Lower energy consumption than Directed Diffusion, even though Directed Diffusion does not run any security primitives. TTSR achieves better routing performance by utilizing powerful H-sensors.

33. Limitations • Due to cost constraints, L-sensors are NOT equipped with tamper-resistant hardware. Assume that if an adversary compromises an L-sensor, she can extract all key material, data, and code stored on that node. • The capability of each individual sensor is not distinguished and the asymmetric links are not fully utilized • When L-sensors try to reach the cluster head, the nodes that are closest to the cluster head have the highest energy burden due to relaying

34. THANK YOU !!

35. Miscellaneous Directed Diffusion • Propagate interest • Set up gradients • Send data and path reinforcement

36. Directed Diffusion • Directed diffusion consists of several elements. Data is named using attribute-value pairs. • A sensing task (or a subtask thereof) is disseminated throughout the sensor network as an interestfor named data. This dissemination sets up gradients within the network designed to "draw" events (i.e., data matching the interest). • Events start flowing towards the originators of interests along multiple paths. The sensor network reinforces one, or a small number of these paths.

37. Directed Diffusion