Wireless Sensor Networks and the Internet of Things: Selected

Wireless Sensor Networks and the Internet of Things: Selected Challenges Delphine Christin, Andreas Reinhardt, Parag S. Mogre, Ralf Steinmetz Multimedia Communications Lab, Technische Universit¨at Darmstadt Merckstr. 25, 64283 Darmstadt, Germany {delphine.christin, andreas.reinhardt, parag.mogre, ralf.steinmetz}@kom.tu-darmstadt.de

II. S ELECTED WSN A PPLICATIONS

Abstract—Wireless sensor networks (WSNs) are increasingly gaining impact in our day to day lives. They are finding a wide range of applications in various domains, including health-care, assisted and enhanced-living scenarios, industrial and production monitoring, control networks, and many other fields. In future, WSNs are expected to be integrated into the “Internet of Things”, where sensor nodes join the Internet dynamically, and use it to collaborate and accomplish their tasks. However, when WSNs become a part of the Internet, we must carefully investigate and analyze the issues involved with this integration. In this paper, we evaluate different approaches to integrate WSNs into the Internet and outline a set of challenges, which we target to address in the near future.

The wide wireless sensor network application field can be divided into three main categories according to [3]: Monitoring space, monitoring objects and monitoring interactions between objects and space. The proposed classification can be extended by an additional category monitoring human beings. One example of the first category is environmental monitoring. WSNs are deployed in particular environments including glaciers [4], forests [3], and mountains [5] in order to gather environmental parameters during long periods. Temperature, moisture or light sensor readings allow analyzing environmental phenomena, such as the influence of climate change on rock fall in permafrost areas [5]. The second category centers on observing particular objects. Structural monitoring is one of the possible illustrations of this category. By sensing modes of vibration, acoustic emissions and responses to stimuli, mechanical modifications of bridges [6] or buildings [7] indicating potential breakages of the structure may be detected. Monitoring interaction between objects and space is the combination of both previous categories and includes monitoring environmental threats like floods [8] and volcanic activities [9]. Presenting an extension to the presented classification, the last category focuses on monitoring human beings. Worn close to the body, the deployed sensors can gather acceleration information and physiological parameters like heart beat rate. Especially in applications in the medical area, such deployments may help diagnosing bipolar patients [10] and monitoring elderly people in a home care scenario [11]. The proposed classification, and particularly the selected deployments, illustrate the high diversity of WSN applications in term of monitored subjects and environments. Beneficial for the Internet of Things, this important scenario diversity must however be taken into account by considering suitable approaches for the WSN integration into the Internet.

I. I NTRODUCTION The future Internet, designed as an “Internet of Things” is foreseen to be “a world-wide network of interconnected objects uniquely addressable, based on standard communication protocols” [1]. Identified by a unique address, any object including computers, sensors, RFID tags or mobile phones will be able to dynamically join the network, collaborate and cooperate efficiently to achieve different tasks. Including WSNs in such a scenario will open new perspectives. Covering a wide application field, WSNs can play an important role by collecting surrounding context and environment information. However, deploying WSNs configured to access the Internet raises novel challenges, which need to be tackled before taking advantage of the many benefits of such integration. The main contributions of this paper can be summarized as follows: We look at WSNs and the Internet holistically, in line with the vision where WSNs will be a part of an Internet of Things. Thereby, we identify representative application scenarios for WSNs (see Section II) from the multidimensional WSN design space [2], in order to obtain insights into issues involved with the integration. These representative application scenarios open up different schemes for integrating the WSNs into the Internet, which we present and compare in Section III. A closer investigation of the integration possibilies then helps us identify critical challenges (see Section IV), which need to be addressed if the full potential of the integration of WSNs and the Internet has to be realized. Finally, in Section V we summarize our discussion, giving pointers for possible solutions to address the identified challenges while regarding the resource limitations present in common WSN nodes.

III. I NTEGRATION A PPROACHES Connecting WSNs to the Internet is possible in the three main approaches mentioned by [12], differing from the WSN integration degree into the Internet structure. Currently adopted by most of the WSNs accessing the Internet, and

31

presenting the highest abstraction between networks, the first proposed approach (Fig. 1) consists of connecting both independent WSN and the Internet through a single gateway.

WSN WSN WSN

G G G

G G G

would break down the connection between both WSN and the Internet. With several gateways and access points, the second and third scenarios do not present such weakness. To ensure network robustness, they would consequently be preferred, if Ensuring network robustness, theyof would the application supports this type networkconsequently structure. be Ensuring network robustness, they would consequently be preferred. The choice between both remaining integration approaches Ensuring they integration would consequently preferred. The choicenetwork betweenrobustness, both remaining approaches be is is influenced by the WSN application scenario. Allowing preferred. The choice between integration approaches is influenced by the both WSNremaining application scenario. Allowing to cover important distances, the second second approach canbe The choiceimportant between both integration approaches isbe influenced by the distances, WSNremaining application scenario. Allowing covering the approach can envisaged for organized mesh topology. Accordinfluencedimportant by the WSN application Allowing covering distances, theinsecond approach can be envisaged forWSNs WSNs organized inscenario. mesh topology. covering distances, second can ingly, thisimportant approach would betheparticularly adapted to be deenvisaged for organized in approach mesh topology. Accordingly, this WSNs approach would be particularly adapted to envisaged for WSNs organized in mesh topology. ployments belonging to the first “Monitoring space” and the Accordingly, this approach be particularly adapted to deployments belonging to thewould first “monitoring space” and the Accordingly, this approach be particularly adapted to second “Monitoring interactions between objects andand space” deployments belonging to thewould firstbetween “monitoring space” the second “monitoring interactions objects and space” deployments belonging to the firstbetween space” and the second “monitoring interactions objects and space” categories previously introduced in the theproposed proposed application categories previously introduced in“monitoring application second “monitoring interactions between objects and space” categories previously introduced the proposed application classification. By Internet ininone-hop, the classification. By offering offering Internetinaccess access one-hop, thethird third categories previously introduced in the application classification. By offering Internet access inWSN one-hop, the requirthird and lastapproach approach be adopted byproposed applications and last cancan be adopted by WSN applications classification. offeringand Internet access in one-hop, the third and last latency approach be adopted WSN applications requiring lowBylatency therefore direct connections. ing low andcan therefore direct by connections. Presenting and last approach can be adopted by WSN applications requiring latency and therefore direct Presenting mainly star WSNs topology, concerned WSNs can mainly starlow topologies, can the maintain suchconnections. organization requiring low latency and by therefore direct connections. Presenting mainly star topology, theof concerned WSNs can conserve such organization having a central gateway by having a central gateway instead a common base station Presenting mainly star topology, the concerned WSNs can conserve such organization by having aInternet central gateway instead of a common base station without access. By without Internet access. By by considering the previous WSN conserve such organization having a central gateway instead of a the common baseWSN station without Internet access. this By considering previous application classification, application this third approach canaccess. be suitable instead of a classification, common station without By considering the previous WSN application classification, this third approach can bebase suitable for objects Internet and human beings for of be object andfor human and [10-11] may considering the can previous WSN application classification, thisbe thirdmonitoring approach suitable objects and human beings monitoring and may be employed in thebeings, [6-7] and employed inand thecan [6],be[7], [10], for [11]objects deployments for example. third approach suitable beings monitoring may be employed in the and [6-7]human and [10-11] deployments for example. monitoring and may be employed in integration thesecond [6-7] and [10-11] It is important tosecond remark both and third indeployments for example. Nevertheless, both andthat third approaches deployments for example. tegration approaches only and support network configuraNevertheless, both second thirdstatic integration support only static network configuration. Indeed,approaches each new Nevertheless, both second and third integration tions. Indeed, each newthedevice wanting to join approaches the supportwanting only static network configuration. Indeed, eachInternet new device to join Internet requires time-consuming support only static network configuration. Indeed, each new device to join the Internet requires time-consuming requires time-consuming gateway reprogramming. Therefore, gatewaywanting reprogramming. Therefore, the flexibility wanted by device wanting to join Internet requires time-consuming gateway reprogramming. Therefore, the flexibility wanted by the flexibility future Internet of the Tings cannot be achieved by cannot both the wanted bythe the future Internet of Things gateway the flexibility wanted by the futurereprogramming. Internet ofapproaches the Therefore, Tings becurrent achieved by both approaches inbytheir current form. cannot be achieved both in their form. the future the Internet of theexpectation, Tings cannot be achieved bytoboth approaches in their current form. ToTofulfill flexibility adopting the “IP the fulfill the flexibility expectation, adopting the “IP to the approaches in their currentexpectation, form. To fulfill the flexibility adopting the “IP to the Field” paradigm [13] be InInthe Field” paradigm [13] may may be appropriate. appropriate. theconsidered considered To fulfill the flexibility expectation, adopting the “IP to the Field” paradigm [13] may be appropriate. In the considered paradigm, sensor nodes are expected to be intelligent network paradigm, sensor [13] nodes are be expected to be In intelligent network Field” paradigm may appropriate. considered paradigm, sensor nodes are be intelligent network components, which will no expected more be to limited tothe sensing tasks. components, which willare no more betolimited to sensing tasks. paradigm, sensor nodes be sensor intelligent network components, which will no expected more be to sensing tasks. By transferring the intelligence to limited the nodes, the By transferring the intelligence to the sensor nodes, the gatecomponents, which will no more be limited to sensing tasks. By transferring the intelligence the sensor nodes, and the gateways functionalities would be to restricted to repetition way’s functionalities would totoreprogramming forwarding and By transferring the Consequently, intelligence to the sensor nodes, and the gateways functionalities wouldbeberestricted restricted repetition protocol translation. gateway protocol translation. Consequently, gateway reprogramming gateways functionalities would be restricted to repetition and protocol Consequently, gateway reprogramming operationstranslation. would no more be required and dynamic network protocol translation. Consequently, gateway reprogramming operations would nobemore more be required required anddynamic dynamic network operations would be and network configuration couldno attained. Additionally, this intelligence operations would nobe more be required and dynamic configuration could beattained. attained. Additionally, thisnetwork shift of configuration could Additionally, this geographicintelligence transfer will open new perspectives including configuration could attained. Additionally, this geographicintelligence intelligence open new perspectives including geographictransfer willwill open new perspectives including based addressing forbe example. transfer will open perspectives including geographicbased addressing for example. based addressing fornew example. NTELLIGENT IV. Ifor based addressing example. SENSOR CHALLENGES CIHALLENGES IV.introduced INTELLIGENT IV. HALLENGES FOR WSN SENSOR IN AN NTERNET T HINGS The C formerly “IP toS the Field” paradigmOFinvolves SENSOR CHALLENGES IV. INTELLIGENT The formerly introduced “IP“IP to to thetheField” involves assigning additional responsibilities to paradigm sensor nodes in The formerly introduced Field” paradigm involves The formerly introduced “IP tofunctionality. the Field” involves assigning additional responsibilities to paradigm sensor nodes in addition toadditional their usual sensing To highlight and assigning responsibilities to sensor nodes in addition assigning additional responsibilities tofrom sensor nodes in addition to their usual sensing functionality. To highlight and discuss the challenges emerging such novel to their usual sensing functionality. To highlight and discuss addition to their usual sensing functionality. To highlight and discuss the assignment, challenges we emerging from potential such novel responsibility selected tasks the challenges emerging fromemerging such novelthree responsibility assigndiscuss the assignment, challenges from such responsibility selected three potential tasks that the sensor nodes wouldwe have to accomplish: securitynovel and ment, selected three potential that the sensor tasks nodes responsibility assignment, we selected three potential that thewe nodes would have totasks accomplish: security and quality ofsensor service management, and network configuration. would to nodes accomplish: Security and configuration. quality of service that thehave would have accomplish: security and quality ofsensor service management, andtonetwork A. Security (QoS) management, and network quality of service management, andconfiguration. network configuration. A. common Security WSNs without Internet access, the sensor nodes In A. Security Security A. In Internet access, maycommon alreadyWSNs playwithout an important role the to sensor ensure nodes data In common without Internet sensor may alreadyWSNs play an important role tothe ensure data confidentiality, integrity, availability andthe authentication In common WSNs without Internetaccess, access, sensornodes nodes may already an important role ensure data confidentiality, integrity, availability andtodata authentication depending onplay theplay application sensitivity. However, the current may already an important role to ensure confidentialconfidentiality, integrity, availability andpresence authentication depending on the application sensitivity. However, the current identified attack scenarios require a physical near ity, integrity, availability and authentication depending onthe the depending on the application sensitivity. the current identified attack require athe physical presence near the targeted WSN inscenarios order to jam, capture orHowever, introduce malicious application sensitivity. However, current identified attack identified attack require a physical presence near the targeted WSN inscenarios orderBy to jam, capture or introduce malicious nodes for example. opening WSNs to scenarios require a physical presence near theInternet, targeted such WSN targeted WSN in order tono jam, capture or introduce nodes example. By opening WSNs to Internet, such locationfor proximity will more be required and malicious attackers in order to jam, capture or introduce malicious nodes nodes forproximity example. By no opening WSNs to Internet, suchfor location will more be required and attackers would be able to threaten WSNs from everywhere. location will no more be everywhere. required and attackers would be proximity able to threaten WSNs from would be able to threaten WSNs from everywhere.

Internet Internet Internet

Gateway Gateway Gateway

Figure Independent network Fig. 1. 1: Independent network Figure 1: Independent network Figure 1: Independent network increasing integration degree,

Showing an the second Showing an increasing integration degree, second Showing(Fig.2) an increasing degree, thethe second apapproach forms a integration hybrid network composed of both Showing an integration degree, the but second approach (Fig.2) forms hybrid network composed of both considered network structure remaining independent, fewof proach (Fig. 2)increasing forms aa hybrid network, still composed approach forms a hybrid network composed ofaccess both considered network remaining independent, but few dual sensor(Fig.2) nodes canstructure access the Internet. independent networks, where few dual sensor nodes can considered network structure remaining independent, but few dual sensor nodes can access the Internet. the Internet. dual sensor nodes can access the Internet. G G G

WSN WSN WSN

G G G

Gateway Gateway Sensor node Gateway Sensor node Sensor node

G G G

Internet Internet Internet

G G G

Figure 2: Hybrid network Fig. 2. 2: Hybrid network Figure Hybrid network Figure Hybrid network the last2: approach is inspired

Illustrated by Fig.3, from current Illustrated by Fig.3, theforms last approach inspired access from current WLAN structure and a denseis802.15.4 point Illustratedbyby Fig.the3,last theapproach last approach is inspired from Illustrated Fig.3, inspired from current WLAN and forms a nodes denseiscan 802.15.4 access point network,structure where multiple sensor join the Internet in current WLAN structure and forms a dense 802.15.4 access WLAN structure and forms a dense 802.15.4 access point network, one hop. where multiple sensor nodes can join the Internet in point network, sensor can join the network, where where multiplemultiple sensor nodes cannodes join the Internet in one hop. Internet one hop.in one hop. WSN WSN WSN G G G

G G G

Gateway Gateway Gateway Gateway Sensor node Gateway Gateway Sensor node Sensor node

G G G

Internet Internet Internet

Figure 3: Access point network Figure 3: Access point network Fig. 3. 3: Access point network network Access point first Figure approach presents a single

However, the point of failure However, first approach presents a single point of failure due to thethe gateway uniqueness. Gateway dysfunction would However, first approach a presents single point ofand failure due thethe gateway uniqueness. Gateway dysfunction would break down the between both WSN the It isto obvious thatconnection the first presents approach a single point due to down the gateway uniqueness. Gateway dysfunction would thetheconnection between both and the Internet networks. With several gateways andWSN access points, ofbreak failure due to gateway uniqueness. Gateway dysfunction break down connection both WSN and the Internet networks. With severalbetween andsuch access points, the second andthethird scenarios dogateways not present weakness. Internet networks. severaldogateways andsuch access points, the second and thirdWith scenarios not present weakness. the second and third scenarios do not present such weakness.

32

example. By opening WSNs to Internet, such location proximity will no more be required and attackers would be able to threaten WSNs from everywhere. In addition to this novel location diversity, WSNs may have to address new threats like malware introduced by the Internet connection and evolving with the attacker creativity. Most current WSNs connected to the Internet are protected by a central and unique powerful gateway ensuring efficient protection. However, a direct reuse of such existing security mechanisms is made impossible by the scarce energy, memory, and computational resources of the sensor nodes. In fact, common Mica2 motes offer 7.3 MHz 8-bit microcontrollers with 128 Kbytes of reprogrammable flash memory, 4 Kbytes of RAM and 4 Kbytes of EEPROM [14]. At last, many services on the Internet make use of cryptography with large key lengths such as RSA-1024, which are not currently supported by sensor nodes. Consequently, innovative security mechanisms must be developed according to the resource constraints to protect WSNs from novel attacks originating from the Internet.

presenting a high diversity in terms of monitored subjects and environments. By taking into account their main characteristics, we have analyzed three integration approaches and demonstrated that they were inappropriate in their current state to allow sensor nodes joining dynamically the Internet of Things. We consider applying the IP to the Field paradigm, which implies assigning additional responsibilities to the sensor nodes as an adequate solution to integrate WSNs with the Internet. We have selected three important task assignments in order to highlight the challenges emerging from the paradigm adoption: Security, QoS, and configuration management. Their analysis revealed that the solutions currently deployed in the Internet are not suitable for the limited sensor node resources and consequently, novel mechanisms have to be developed to adapt to the capabilities and constraints of WSNs. We plan to investigate existing approaches and find suitable modifications for resource-constrained sensor platforms to tackle these challenges.

B. Quality of Service

ACKNOWLEDGMENT This work was supported by CASED (www.cased.de). The authors would like to thank Matthias Hollick for the fruitful discussions.

With gateways acting only as repeater and protocol translators, sensor nodes are also expected to contribute to quality of service management by optimizing the resource utilization of all heterogeneous devices that are part of the future Internet of Things. Not considered as a weakness, the device heterogeneity opens new perspectives in terms of workload distribution. In fact, resource differences may be exploited to share the current workload between nodes offering available resources. Improving the QoS, such collaborative work is consequently promising for mechanisms requiring high amount of resources like security mechanisms. Nevertheless, the existing approaches ensuring QoS in the Internet are not applicable in WSNs, as sudden changes in the link characteristics can lead to significant reconfiguration of the WSN topology. It is therefore mandatory to find novel approaches towards ensuring delay and loss guarantees.

R EFERENCES [1] “Internet of Things in 2020: Roadmap for the Future,” 2008, online, http://www.smart-systems-integration.org/public/internet-of-things. [2] K. R¨omer and F. Mattern, “The Design Space of Wireless Sensor Networks,” Wireless Communications, IEEE, vol. 11, no. 6, 2004. [3] D. Culler, D. Estrin, and M. Srivastava, “Guest Editors’ Introduction: Overview of Sensor Networks,” vol. 37, no. 8, 2004. [4] K. Martinez, R. Ong, and J. Hart, “Glacsweb: a Sensor Network for Hostile Environments,” in Proceedings of the Sensor and Ad Hoc Communications and Networks Conference (SECON), 2004. [5] I. Talzi, A. Hasler, S. Gruber, and C. Tschudin, “PermaSense: investigating permafrost with a WSN in the Swiss Alps,” in Proceedings of the workshop on Embedded networked sensors (EmNets), 2007. [6] R. Lee, K. Chen, S. Chiang, C. Lai, H. Liu, and M.-S. Wei, “A Backup Routing with Wireless Sensor Network for Bridge Monitoring System,” in Proceedings of the Communication Networks and Services Research Conference(CNSR), 2006. [7] P. Katsikogiannis, E. Zervas, and G. Kaltsas, “A Wireless Sensor Network for Building Structural Health Monitoring and Seismic Detection,” physica status solidi (c), vol. 5, 2008. [8] D. Hughes, G. Blair, G. Coulson, P. Greenwood, B. Porter, P. Smith, and K. Beven, “An Adaptable WSN-based Flood Monitoring System,” in Proceedings of the European Conference on Smart Sensing and Context (EuroSSC), 2007. [9] W. Werner-Allen, K. Lorincz, M. Ruiz, O. Marcillo, J. Johnson, J. Lees, and M. Welsh, “Deploying a Wireless Sensor Network on an Active Volcano,” IEEE Internet Computing, vol. 10, no. 2, 2006. [10] M. H. Teicher, “Actigraphy and Motion Analysis: New Tools for Psychiatry,” Harvard Review of Psychiatry, vol. 3, 1995. [11] A. Wood, G. Virone, T. Doan, Q. Cao, L. Selavo, Y. Wu, L. Fang, Z. He, S. Lin, and J. Stankovic, “ALARM-NET: Wireless Sensor Networks for Assisted-living and Residential Monitoring,” 2006. [12] R. Roman and J. Lopez, “Integrating Wireless Sensor Networks and the Internet: a Security Analysis,” Internet Research: Electronic Networking Applications and Policy, vol. 19, no. 2, 2009. [13] Smart Energy Alliance, online, http://www.smart-energyalliance.com/solutions/ip-to-the-field/. [14] Crossbow Technology, online, http://www.xbow.com.

C. Configuration In addition to security and QoS management, sensor nodes can also be required to control the WSN configuration, which includes covering different tasks, such as address administration to ensure scalable network constructions and ensuring self-healing capabilities by detecting and eliminating faulty nodes or managing their own configuration. However, selfconfiguration of participating nodes is not a common feature in the Internet. Instead, the user is expected to install applications and recover the system from crashes. In contrast, the unattended operation of autonomous sensor nodes requires novel means of network configuration and management. V. C ONCLUSION In this first analysis step to integrate WSNs into the Internet of Things, we have considered selected application scenarios

33