You are on page 1of 6

A Secure Self-Organizing Sensor Network

Dan C. Marinescu, Chen Yu, and Gabriela M. Marinescu School of Electrical Engineering & Computer Science University of Central Florida, Orlando, FL 32816, USA Email: (dcm, yuchen, magda)@cs.ucf.edu Abstract
In this paper we discuss a scheme that allows sensor networks to operate securely and efciently. In this scheme after deployment the sensors go through a self-organization phase when they establish a communication pattern among themselves; then they follow this pattern through multiple activity phases when they collect, process, and transmit information. The algorithm for self-organization assumes anonymous sensors and random times of the communication events, as well as random communication frequencies; the sensors use random number generators and a set of shared seeds so no external entity can either join in, or predict the time when the sensors in the set will transmit and attempt to interfere with the transmissions. The scheme extends the lifetime of the network by reducing the power consumption; it minimizes the number of collisions experienced by a sensor when it transmits and maximizes the time a sensor is either idle or dedicated to monitoring and/or internal data processing.

1 Motivation and Related Work


A sensor network consists of spatially distributed autonomous devices equipped with a radio transceiver or other wireless communication device, a micro-controller, an energy source, and sensors which monitor temperature, sound, vibration, pressure, motion, chemical pollutants, radiation, or other physical characteristics of the environment. The low-cost nodes of a wireless sensor network are anonymous and have limited resources, CPU cycles, memory, power, communication hardware and range, thus they have to collaborate to accomplish any meaningful task. Information assurance is an important attribute of a sensor network, especially in a hostile environment; it refers to availability, integrity, authentication, condentiality, and no-repudiation. Authentication means the ability to ensure that each sensor is indeed a legitimate one, no-repudiation

means that each sensor will keep its end of the deal once it enters a contract. Efcient use of resources and information assurance are primordial considerations that should be addressed at the time when the system is designed, they cannot be an add on at a later time. Energy awareness for networks of mobile devices has been investigated for some time [2, 4, 6, 10, 11, 12, 17, 18, 24, 25, 26, 28, 29, 32, 33, 35]. Some of the proposals are biologically inspired, e.g., the behavior of a node is controlled by a tness function, the node communicates only if the evaluation of the tness function leads to a value exceeding a threshold [28], others are inspired by social systems [4]. Distributed computing in wireless network has also been proposed [18]. Multiple aspects of the organization and operation of sensor networks based upon a range of assumptions have been reported in the literature in recent years [1, 9, 14, 23, 27]. The research has been focused on energy efcient physical layer protocols or Medium Access Control (MAC) protocols [2, 24, 26, 30, 35], routing and topology maintenance [4, 10, 11, 15, 28, 29, 32, 33], coordination, synchronization, and information dissemination [6, 16, 36], information assurance [8, 13, 21, 22], and reliability [31]. Some of the algorithms and protocols discussed in the literature attempt to adapt known techniques from traditional wireless networks to sensor networks, others reect the view that the traditional collaboration and communication schemes cannot be easily applied to a sensor network. Traditional collaboration schemes require precise identication of each entity, unfeasible for anonymous sensors, and mechanisms to ensure some form of fault-tolerance that require constant monitoring of the system thus, substantial power consumption. The trust management schemes we are aware of suffer from similar problems; encryption and decryption are CPU-intensive activities likely to drain the power reserves of the sensors rather fast. Even the routing and MAC protocols developed for wireless networks require a unique id of the individual nodes and cannot be immediately applied to sensor networks with anonymous nodes.

We propose to combine a self-organization strategy that guarantees scalability and minimizes the energy consumption of the sensor network with a scheme for sharing genetic information that guarantees information assurance. The sensors rst go through a self-organization phase when they establish a communication pattern among themselves; then they follow this pattern through multiple activity phases when they collect, process, and transmit information. The algorithm for self-organization assumes anonymous sensors and random times of the communication events, as well as random communication frequencies; the sensors use random number generators and a set of shared seeds so no external entity can either join in, or predict the time when the sensors in the set will transmit and attempt to interfere with the transmissions. The scheme proposed in this paper extends the lifetime of the network by reducing the power consumption; it minimizes the number of collisions experienced by a sensor when it transmits and maximizes the time a sensor is either idle or dedicated to monitoring and/or internal data processing. The scheme can be extended to allow a sensor network to adapt to topological changes, include new batches of sensors, and disregard the sensors that have depleted their power reserves, or have been compromised.

2 The Model of a Scale-free Sensor Networks


The property of a system to be scale-free is critical in an environment with limited resources. The scheme we propose limits the number of partners each sensor collaborates with, thus, it limits the amount of communication and the complexity of coordination. The scheme ensures scalability, the number of sensors each sensor communicates with is limited regardless of the total number of sensors in the network; a node in the random graph showing the systems connectivity is linked with at most other nodes. The existence of multiple paths among pairs of nodes increases the redundancy, thus it increases the chances of the system to survive. The sensors are anonymous, but during our discussion we shall identify a sensor by a name, e.g., i , to facilitate the description of the algorithm. During the self-organization phase a sensor assumes a pseudo-Id, I the index of the rst slot it is able to successfully transmit, as discussed in Section 3. The PseudoId is guaranteed to be unique and allows a sensor to limit the number of sensors it communicates with. Sensors are deployed in batches. The basic algorithm discussed in this paper considers a single batch. The scheme we propose is based upon the following assumptions: (i) The time is slotted and the duration of a slot is chosen to allow each sensor to complete its transmission during the

activity phase. During the self-organization phase each slot is in turn divided into micro-slots. During a micro-slot a sensor could transmit only a short packet. This short packet could be a request to reserve a micro-slot, or a short message. (ii) The lifetime of the sensor network is divided into epochs Ep , 1 p. For the basic algorithm we assume that all epochs consists of the same number of events, . Each epoch consists of a self-organization phase followed by activity phases. All phases of an epoch consist of the same number of events. Thus, the number of events in an epoch is = ( + 1). (iii) Each sensor has built in genetic material including random number generators and seeds q , 1 q 3 as follows: 1 is the seed to determine a random sequence of time slots ti ; 2 is the seed to determine a random sequence of frequency channels i ; and 3 is the seed to determine i a random hoping frequency j for each channel frequency i . A sensor is able to determine its location; this is not a critical requirement of the algorithm, but it is useful to establish sensor proximity. In absence of this facility the sensors use the signal strength to determine the ones in their proximity. (iv) The sensors have a limited range and the density of the sensors is , in other words in average we expect to have sensors per square unit of the area covered by the sensor network. Each sensor a is thus able to construct its proximity set, Pa consisting of sensors in its vicinity that will wake-up during the activity phases and receive the information sent by a. (v) The network is dense; this means that = 2 , the average number of sensors that are able to receive the transmission of a sensor is at least p with p a small integer, 4 p 6 and the cardinality of the proximity sets. (vi) The sensors in a batch have their clocks synchronized to its sink . A sink is a sensor with special properties: (a) Links the sensor network with the outside world; it communicates with an external controller (a satellite, a drone, an aircraft, or even a stationary device) to report relevant information. (b) Has a larger power reserve and a transmission range that covers the entire deployment area. All sensors synchronize their clock to the master clock of at the beginning and at the end of each phase. (c) The genetic material of the sink, in addition to the elements described at (iii) also includes two integers whose role will be claried shortly, , the number of events in one phase, , the number of activity phases, as well as , the cardinality of the proximity set, M the cardinality of the event index set Q used to keep track of transmission and receiving events, and , a parameter for synchronization; the sink broadcasts , , , M , and at the beginning of the self-organization phase. (d) The sink transmits at full power only during the

rst and the last slot of a phase (self-organization or activity phases) when it disseminates information, and/or provides a master clock for synchronization. During the rest of a phase it operates like any other sensor and has the same range as all the other sensors. For now we assume a unique sink, but fault-tolerance requires backup sinks that can take over if the original sink fails. (vii) The energy dissipation for a transmission during the t self-organization phase is Eo and during the activity phase t t is Ea > Eo ; the energy dissipation when a sensor wakes r and up and listens during the self-organization phase is Eo r t during the activity phase is Ea > Eo . (viii) A sequence of events 1 , 2 , 3 , . . . , k1 , k , k+1 , . . . , that occur in slots starting at random times t1 < t2 < t3 . . . < ti1 < ti < ti+1 . . . control the time evolution of the system. An event has a global index, an index within an epoch, and an index within a phase. For the sake of simplicity in the description of the algorithm instead of the slot corresponding to the event k we say slot of k ; similarly instead of the rst micro-slot of the slot corresponding to the event k we say the rst micro-slot of slot k . (ix) The senors are autonomous in terms of data collection and local processing, but reactive in terms of communication. During the self-organization phase the sink transmits during the rst event. A sensor only responds to a successful transmission of another sensor after evaluating two tness functions f, g [28]; only if the value of the tness function exceeds a certain threshold the sensor is allowed to transmit. This threshold is function of a number of parameters including the strength of the incoming signal, the power reserve of the sensor, and the location information if available. During the activity phase the slots when sensors wake up to transmit or receive correspond to scheduled events; this schedule is established during the self-organization phase. The construction of the tness function and the determination of the threshold are fairly complex subjects and are not discussed in this paper. (xi) A sensor including dwells micro-seconds on each frequency in hopping sequence. When a sensor wakes up for the event i its master clock is in one of the time slots ti1 , ti , or ti+1 . Each sensor knows the last frequencies i1 , i , and i+1 on which the sender will dwell in the time slots ti1 , ti , and ti+1 . A strategy to allow synchronization in the presence of clock drift is: tune in, cyclically, to i1 , i , and i+1 spending /3 time units on each of them. (xii) The event i , a transmission with frequency i , occurs at the beginning of the slot with index i, at time ti . The index i of the event i is a global pointer into a random sequence of frequencies and time slots.

3 The self-organization Phase


Our goal is to design a scale-free sensor network; in this network each sensor is connected with a limited number of other sensors. In the context of this paper the term connected means that a node has a number of neighbors and is able to construct a proximity set P of sensors it communicates with. Pi , the proximity set of sensor i consists of at most sensors i is aware of. The sensors in Pi are able to receive transmissions initiated by i because all sensors have the same transmission range. During the self-organization phase a sensor assumes a pseudo-identity, I given by the index of the rst event when it has transmitted successfully; for example, the sensor we call i will assume the identity Ii = k if i successfully transmits during the slot corresponding to the event k . Before the rst successful transmission the PseudoId of each sensor is set to zero and then reset to the index of the slot when the successful transmission takes place. This PseudoId is stamped on every message sent during the selforganization phase to guarantee that the proximity set of a sensor does not grow beyond the limit imposed, | P | . The PseudoId is available during the activity phase yet we do not see any advantage to include it in messages sent during the activity phases. When i includes a sensor with PseudoId Isender in its proximity set it also creates a counter of events when it has received messages from it, counter(Isender ) as well as an event index list Event(Isender ), of size at most M . It follows that the storage requirements for each sensor are: (2 + M ) integers. During the self-organization phase there are two types of messages: ReqToJoin (RtJ) and ReqToForward (RtF). The rst signals to other sensors that the sender is looking for other sensors to join its proximity set; the message always includes the PseudoId of the sender. The second type of message signals that a sender cannot accept an invitation to join and simply forwards it to others to keep the organization process going. A ReqToForward always contains the PseudoId of the originator of a ReqToJoin message, rather than the PseudoId of the sensor forwarding the message. The role of the self-organization phase is to: (a) Allow all sensors to construct their proximity sets; (b) Establish the communication pattern between a sensor and the members of its proximity set during all activity slots. At the conclusion of the self-organization phase each sensor i knows the index k of the event k Pi when it is scheduled to wake-up; during these events the members of its proximity set, Pi , are scheduled to wake-up and receive the transmission. Sensor i uses two tness functions: fi to determine if it should respond with a ReqT oJoin message and gi to determine if it should respond with a ReqT oF orward message to a successful transmission of another sensor.

We stress again that knowing the index k event k does not mean that we know the time tk of the occurrence of the event. These times tk are random, but all sensors can determine them as they share the seeds used by the random number generators used to determine them. The self-organization algorithm followed by sensor i with proximity set Pi and tness functions fi and gi when i observes a successful transmission in the slot of k from a sensor with PseudoId Isender is shown in Figure 1.

5 Conclusions and future work


The algorithm presented in this paper supports freescale, self-organizing sensor networks. Self-organization is a trait of adaptive systems with limited resources, it ensures scalability and confers the ability to survive. During a self-organization phase we construct a random graph connecting a sensor with a bounded number of other sensors. Each sensor a builds a proximity set, Pa with at most members. For each member of the proximity set senor a stores at most M event indexes; these events are used during the following activity phases to determine when sensor a should wake up to transmit and receive information from the members of Pa . The construction of the proximity sets of bounded cardinality, Pa is a rather challenging task when one assumes that the sensors are anonymous. We use the collision resolution algorithm to attribute a unique PseudoId to each sensor and thus limit the number of sensors in a proximity set; this PseudoId plays no role during the activity phase when sensors are only aware of the indices of the slots when they are expected to wake up and either send or receive a message. The communication schedule followed by all sensors prevents collisions during the activity phases and minimizes the time when the sensors use their RF sub-system to send or receive messages; thus, it minimizes the power consumption of individual sensors and extends the lifetime of the sensor network. The system provides information assurance with minimal power consumption and avoids expensive encryption/decryption techniques discussed in the literature; the time and the frequency used for communication are random numbers and no external entity can either join in, or predict the time when the sensors in the set will transmit and attempt to interfere with the transmissions. The algorithm is intricate, but requires only limited resources. The intricacy of the algorithm is a consequence of the desire to support self-organization, a media access control that minimizes the number of collisions, and the commitment to information assurance. An analysis of the algorithm can be difcult, but simulation studies to verify the feasibility of the scheme discussed in this paper have shown that the scheme works well.

4 Activity Phase
During an activity phase the sensors carry out their monitoring function and report partial results. The schedule of events for each sensor, namely the slots when it transmitted successfully during the self-organization phase are known, and this knowledge allows each sensor to determine the slots during the activity phase they have to wake-up to receive information and then transmit. The sink transmits in the rst micro-slot of 1 then i transmits in one of the micro-slots of 2 , j transmits in one of the micro-slots of 3 , m transmits in one of the micro-slots of k , n transmits in one of the micro-slots of 1 , and nally the sink transmits in the rst micro-slot of . The event in an activity phase labeled as k is the reciprocal of the event k1 from the self-organization phase in the following sense: if sensor x was successful to transmit in slot k during the self-organization phase then it will be scheduled to wake-up and receive in slot k 1 = k and then transmit in slot k = k + 1. The algorithm followed by sensor i is: (a)Determine the time of rst event of the activity phase and wake up to synchronize the clock to the sink . (b) Compute the reciprocal events for all events included in the EventIndexList for all members of the proximity set Pi . (c) Construct an ordered list of all the indices of reciprocal events. If the index of the last event recoded by i is klast , then the index of the reciprocal event is kf irst = klast = klast + 1. It follows that i must rst wake up in slot klast 1 = k to receive a transmission and then transmit in slot klast = k + 1. (d) Determine the time of the next event when the sensor must wake up. If the next reciprocal event is k then determine the time of the event k 1; wake up and receive at that time then transmit in the next slot. (e) Repeat the previous step until all reciprocal events, at most M , have been exhausted. (f) Wait until signals the end of the activity phase. Ea , the power dissipated by a sensor for coordination during an activity phase is upper bounded: Ea t r M (Ea + Ea ).

6 Acknowledgements
This research was supported in part by National Science Foundation grants ACI0296035, EIA0296179 and by a 2007 Ernest Thomas Sinton Walton award from the Science Foundation of Ireland.

Figure 1. The Self-Organization Algorithm.

References
[1] I. Akyildiz, W. Su, Y. Snakarasubranmaniam, and E. Cayirci. A Survey on Sensor Networks. Computer Networks, 38: 393-422, 2002. [2] K. Arisha, M. Youssef, and M. Younis. Energy Aware TDMA Based MAC for Sensor Networks. Proc. IEEE Integrated Management of Power Aware Communications, Computing and Networking (IMPACCT 2002), New York, May 2002. [3] E. Bonabeau, M. Dorigo, and G. Theraulaz. Inspiration from Optimization from Social Insect Behavior. Nature, 406: 39-42, 2000. [4] D. Braginsky and D. Estrin. Rumor Routing Algorithm for Sensor Networks. Proc. 1st ACM International Workshop on Wireless Sensor Networks and Applications, pp.22-31, ACM Press, 2002. [5] S. F. Camazine, J.-L. Deneubourg, N. R. Franks, J. Sneyd, G. Theraulaz, and E. Bonabeau. Self-Organization in Biological Systems., Priceton University Press, Princeton, NJ, 2001.

[6] B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris. SPAN: An Energy Efcient Coordination Algorithm for Topology Maintenance in Ad Hoc Wireless Networks. ACM Wireless Networks Journal, 8(5): 481 - 494, 2002. [7] T. C. Collier and C. Taylor. Self-Organization in Sensor Networks. Journal of Parallel and Distributed Computing, 64(7): 866 - 873, 2004. [8] B. Deb, S. Bhatnagar, and B. Nath. Information Assurance in Wireless Sensor Networks. Proc. Wireless Sensor Networks and Applications (WSNA03), San Diego, CA, 2003. [9] D. Estrin, M. Srivastava, and A. Sayeed, Wireless Sensor Networks. 8th Int. Conf. on Mobile Computing and Networking (MobiCom02), Atlanta, GA, 2002. [10] D. Ganesan, R. Govindan, S. Shenker, and D. Estrin. Highly Resilient, Energy Efcient Multipath Routing in Wireless Sensor Networks. Mobile Computing and Communications Review (MC2R), vol. 1, No. 2, 2002. [11] T. He, J. Stankovic, C. Lu, and T. Abdelzaher. SPEED: A Real Time Routing Protocol for Sensor Networks. Proc. Int. Conf. on Distributed Computing Systems, p.46, Providence, RI, 2003.

[12] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan. Energy Efcient Communication Protocol for Wireless Microsensor Networks. Proc. Hawaii Int. Conf. System Sciences, Hawaii, 2000. [13] Q. Huang, J. Cukier, H. Kobayashi, and J. Zhang. Fast Authentication Key Establishment Protocols for SelfOrganizing Sensor Networks. Proc. Wireless Sensor Networks and Applications (WSNA03), San Diego, CA, 2003. [14] C. Intanagonwiwat, R. Govindan, and D. Estrin. Directed Diffusion: A Scalable and Robust Communication Paradigm for Sensor Networks. Proc. 6th Int. Conf. on Mobile Computing and Networking (MobiCom00), Boston, MA, 2000. [15] C. Karlof, Y. Li, and J. Polastre. ARRIVE: Algorithm for Robust Routing in Volatile Environments. Technical Report UCB/CSD-03-1233, U. C. Berkeley, May 2002. [16] J. Kulik, W. Rabiner, and H. Balakrishnan. Adaptive Protocols for Information Dissemination in Wireless Sensor Networks. Proc. 5th Int. Conf. on Mobile Computing and Networking (MobiCom99), Seattle, WA, 1999. [17] Q. Li, J. Aslam, and D. Rus. Hierarchical Power Aware Routing in Sensor Networks. The DIMACS Workshop on Pervasive Networking, May 2001. ol oni, and [18] D.C. Marinescu, G. M. Marinescu, Y. Ji, L. B H. J. Siegel. Ad Hoc Grids: Communication and Computing in a Power Constrained Environment, Workshop on Energy-Efcient Wireless Communications and Networks 2003 (EWCN 2003), Proc. 22nd Int. Performance, Computing, and Communications Conf. (IPCCC), IEEE Press, pp. 113122, 2003. [19] D. C. Marinescu, C. Yu, G. M. Marinescu, J. P. Morrison, and C. Norvik. A Reputation Algorithm for a Self-organizing System Based upon Resource Virtualization, Proc. 22nd IEEE Int. Parallel and Distributed Processing Symposium (IPDPS08), Heterogeneity in Computing Workshop 2008 (HCW-08), CD Procedings, 2008. [20] D. C. Marinescu, C. Yu, and G. M. Marinescu. Selforganizing Sensor Networks. Proc. Int. Symp. on Wireless Pervasive Computing (ISWPC 2008), CD Proceedings, 2008. [21] S. Olariu and Q. Xu. Information Assurance in Wireless Sensor Networks. Proc. Wireless Sensor Networks and Applications (WSNA03), San Diego, CA, 2003. [22] S. Olariu. Information Security in Wireless Sensor Networks. http://i2.lab.ucf.edu, 2007. [23] S. Park, A. Savvides and M. Srivastava. Simulating Networks of Wireless Sensors. Proc. 33rd Winter Simulation Conf., pp. 1330-1338, Arlington, VA, 2001. [24] C. Raghavendra and S. Singh. PAMAS: Power Aware Multi Access Protocol with Signaling for Ad Hoc Networks. ACM SIGCOMM Computer Communication Review, 28(3): 5 - 26, 1998. [25] V. Raghunathan, C. Schurgers, S. Park, and M. Srivasatava. Energy Aware Wireless Microsensor Networks. IEEE Signal Processing Magazine, 19(2): 40-50, 2002.

[26] E. Shih, S. Cho, N. Ickes, R. Min, A. Sinha, A. Wang, and A. Chandrakasan. Physical Layer Driven Protocol and Algorithm Design for Energy Efcient Wireless Sensor Networks. Proc. 7th Int. Conf. on Mobile Computing and Networking (Mobicom01), Rome, Italy, 2001. [27] K. Sohrabi, J. Gao, V. Ailawadhi, and G. Pottie. Protocols for Self-organisation of a Wireless Sensor Network. IEEE Personal Communications, 7(5): 16-27, 2000. [28] G. Wang, D. Turgut, L. B ol oni, Y. Ji, and D.C. Marinescu. Improving Routing Performance Through m-Limited Forwarding in Power-Constrained Wireless Networks. Journal of Parallel and Distributed Computing (JPDC), Vol. 68, pp. 501514, 2008. [29] M. Woo, S Singh, and C. Raghavendra. Power Aware Routing in Mobile Ad Hoc Networks. Proc. 4th Annual ACM/IEEE Int. Conf. on Mobile Computing and Networking (MobiCom98), Dallas, TX, 1998. [30] A. Woo and D. Culler. A Transmission Control Scheme for Media Access in Sensor Networks. Proc. 7th Int. Conf. on Mobile Computing and Networking (MobiCom01), pp. 221 - 235, 2001. [31] A. Woo. Evaluation of Efcient Link Reliability Estimators for Low Power Wireless Networks. Technical Report No. UCB/CSD-03-1270, U. C. Berkeley, 2003. [32] Y. Xu, J. Heidemann, and D. Estrin. Geography-informed Energy Conservation for Ad Hoc Routing. Proc. 7th Int. Conf. on Mobile Computing and Networking (Mobicom01), Rome, Italy, 2001. [33] Y. Yu, R. Govindan, and D. Estrin. Geographical and Energy Aware Routing: A Recursive Data Dissemination Protocol for Wireless Sensor Networks. Technical Report, UCLA-CSD TRE-01-0023, UCLA, May 2001. [34] F. Ye, S. Lu, and L. Zhang. GRAdient Broadcast: A Robust, Long-lived Large Sensor Network. Technical Report, http://irl.cs.ucla.edu/papers/grab-tech-report.ps, 2001. [35] W. Ye, J. Heidemann, and D. Estrin. An Energy Efcient MAC protocol for Wireless Sensor Networks . Proc. IEEE Infocom 2002. New York, June 2002. [36] F. Ye, H. Luo, J. Cheng, S. Lu, and L. Zhang. A Two-Tier Data Dissemination Model for Large Scale Wireless Sensor Networks. Proc. 8th Int. Conf. on Mobile Computing and Networking (Mobicom02), Atlanta, GA, 2002.