View
4
Download
0
Category
Preview:
Citation preview
Self-Organization in Autonomous Sensor/Actuator
Networks
Falko DresslerAutonomic Networking GroupDept. of Computer Sciences
University of Erlangendressler@informatik.uni-erlangen.de
Bio-inspiredNetworkingBio-inspiredNetworking
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
2
Outline
Autonomic Networking GroupIntroductionResearch projects
Selected research issuesBio-inspired networking
Feedback-loopCongestion control
ProfilingRobot-assisted WSN
Data communication in WSN(semi-)reliable and authenticated communication in WSN
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
3
Autonomic Networking Group
Dept. of Computer SciencesComputer Networks and Communication Systems
Autonomic Networking Group
Group membersFalko Dressler (group leader)Thomas Halva Labella (guest researcher)Isabel Dietrich (Ph.D. candidate)Gerhard Fuchs (Ph.D. candidate)… and numerous B.Sc. and M.Sc. candidates
Research objectivesAutonomous Sensor/Actuator NetworksBio-inspired NetworkingNetwork Monitoring and Security
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
4
Autonomous Sensor/Actuator Networks
Sensor networksSensor motes: sensors + processing + wireless communication + batterySensor networks: ad hoc networks consisting of hundreds of sensor motesApplication: logistic, security, environmental observations, health, home automation, pervasive computingResearch issues: addressing, ad hoc routing, group communication, task allocation, coordination, self-organization, energy efficiency, time synchronization, coverage, localization, security, quality of service
Mobile robotsCooperating mobile autonomous systemsApplication: security, service, entertainmentResearch issues: localization, navigation, cooperation,coordination, tracking
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
5
ROSES
RObot assisted SEnsornetworkS
Employment of robot systems forMaintenance in sensor networkProviding communication relays
Sensor assisted teams of robotsMore accurate localizationNavigationInfrastructure forenhancedcommunication
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
6
ROSES
Research GoalsEnergy efficient operation, communication, and navigationSensor network assisted localization and navigation of the robotsUtilization of the robots as a communication relaybetween a sensor network and a global networkQuality of service aware communication in heterogeneous mobile networks with dynamic topologyOptimized task allocation and communication based on application and energy constraintsSecure communication and data management in mobile sensor networks
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
7
BioNeting - Bio-inspired Networking
ConceptsMapping of mechanisms from cellular and molecular biology to networking architecturesStudy of large scale networksAnalyzing the internal functions of network components as well as their interactions in comparison with cellular systems and the associated intra- and inter-cellular signaling pathways
Research GoalsAnalysis of similarities of computer networks andcellular systemsDeduction of new concepts for behavior patterns of network nodesIncreasing the efficiency of individual subsystems
Bio-inspiredNetworkingBio-inspiredNetworking
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
8
Outline
Autonomic Networking GroupIntroductionResearch projects
Selected research issuesBio-inspired networking
Feedback-loopCongestion control
ProfilingRobot-assisted WSN
Data communication in WSN(semi-)reliable and authenticated communication in WSN
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
9
Self-Organization in Pervasive Environments
Identification of availableCommunication pathsNodesCapabilitiesResources
Handling of dataStorageAggregationDistribution
Without knowledge about topology, available nodes, their addresses, their location, …
S
SSSS
Requestfor sensor
data
Local information exchangeRelayed requests
S
SS
SS
SS
SSSS
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
10
Biological Information Exchange
DNA
Signal(information)
Gene transcriptionresults in the formation of a specific cellular response to the signal
Receptor
DNA
Signal(information)
Gene transcriptionresults in the formation of a specific cellular response to the signal
Receptor
DNATissue 1
Tissue 2
DNA
DNA
DNA
DNA
DNA
Tissue 3
Blood
DNADNATissue 1
Tissue 2
DNADNA
DNADNA
DNADNA
DNADNA
DNADNA
Tissue 3
Blood
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
11
Adaptation to Networking
Lessons to learn from biologyEfficient response to a requestShortening of information pathwaysDirecting of messages to an applicable destination
Application for efficient, i.e. optimizedTask allocationInformation exchangeGroup communication
Example: feedback loop mechanismdata dissemination and control loop in ad hoc sensor networks + self-initiated adaptive congestion controlSensor communication actuator
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
12
Excurse: Current Solution
Occurrence of a critical eventTemperature dropsA sensor node detects the problemA request for solving the task is initiated
Start of the reactionThe request reaches an node which can solve the problemThe countermeasure is taken
2
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
13
Excurse: Current Solution
Confirmation of the taskA confirmation message is sent back to the initiator of the requestFor this state information is accumulated in several nodes
Arrival of the responseThe confirmation message receives the initiatorThe transaction is finished
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
14
Excurse: Regulation of Blood pressure
+
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
15
Shifting the Paradigms
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
16
Shifting the Paradigms
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
17
Shifting the Paradigms
request
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
18
Shifting the Paradigms
The smooth muscle cells, the kidney and the brain team up one “meta” nodeThis node knows the answer to the request
request
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
19
Shifting the Paradigms
No confirmation message is neededThe change of the environment indicates the successful initiation of the task
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
20
Feedback Loop
Bio-inspired networking in generalComparing cellular structures with computer networks, we always find similar structuresSpecific signaling pathways are directly adaptable to information exchange in sensor networks (and to other networking issues)Self-organization of complex operations becomes possible using bio-inspired mechanisms
Feedback loop mechanismEmployment of out-of-band methods for task managementSuppression techniques for congestion-aware signalingReduction of control messagesUsage of asymmetric information pathwaysAdaptive congestion control
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
21
Outline
Autonomic Networking GroupIntroductionResearch projects
Selected research issuesBio-inspired networking
Feedback-loopCongestion control
ProfilingRobot-assisted WSN
Data communication in WSN(semi-)reliable and authenticated communication in WSN
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
22
Data Dissemination in SN
Assumptions for typical sensor networksLarge number of participating nodesEach node transmits its measurement data with the same constant data rateDuring fast environmental changes ( emergencies), this data rate may show high bursts
Data disseminationAddress centric routing and data forwarding
due to scalability issues infrequently used in SNData-centric message forwarding, e.g. flooding, gossiping, probabilistic/weighted forwarding, diffusion techniques
usually lack congestion control mechanisms
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
23
Congestion Control in SN
State of the artEnd-to-end – flow based, relying on window mechanisms or feedback informationPath-based – flow based, using congestion signaling along the data path towards the senderHop-by-hop – detection information is signaled to all neighboring nodes
All approaches are based on Internet technologyrequire unambiguous addressing informationrequire defined data paths between sender and receiver
Few of these requirements can be granted in typical wireless sensor networks
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
24
Reasons for Congestion in SN
GW20°C
21°C
22°C
21°C
20°C
21°C
(A)
Congestion due to changes in the environment
Similar or even worse situations can be expected in heterogeneous network environments employing different sensors for multiple simultaneous tasks!
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
25
Congestion Behavior
time
# m
essg
aes
lowhighlow+highlink capacity
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
26
First Enhancements
Weighted probabilistic data dissemination1. assign importance I to each event2. calculate priority P(I) describing the distribution range3. for all neighboring nodes Nn and previously known remote
accessible nodes Nr, calculate an exponentially distributed weighting W(N)
4. forward message if W(N)<P(I)
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
27
Locality Driven Congestion Control
Basic requirementsTo maintain control even if some links get temporarily saturatedTo give priority to important messagesTo prevent starvation of particular transmissions
AlgorithmBased on the number of successfully received messages N during the last time interval T
For each message Mupdate message counter N(M,T)
identify importance factor IMcalculate probability P(N,IM)if exponentialDist(P,T)=TRUE
forward message M
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
28
Simulation Results
time
# m
essg
aes
lowhighlow+highlink capacity
time
# m
essg
aes
lowhighlow+highlink capacity
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
29
Locality Driven Congestion Control
General requirements and solution spaceSelf-organizing and adaptive communication mechanisms are required for large-scale sensor networksStudies of biological mechanisms for self-organization such as the signaling pathways in cell and molecular biology provide high potentials
Locality driven congestion control provides the following features
Based on locally available information onlyhighly scalableno further control overhead / message overhead
Adaptive to changing network conditionsFeatures priority messagesNo starvation of less important transmissions
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
30
Outline
Autonomic Networking GroupIntroductionResearch projects
Selected research issuesBio-inspired networking
Feedback-loopCongestion control
ProfilingRobot-assisted WSN
Data communication in WSN(semi-)reliable and authenticated communication in WSN
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
31
Profiling
MotivationConsider a stationary WSN consisting of possibly heterogeneous sensorsMultiple different applications exist to be executed by the sensorsSensors have different capabilities; different versions of the software might existDeployment of large-scale WSN is difficult, maintenance even more
GoalsMobile robots are used for maintenance of WSN
ConfigurationReconfiguration(Re-)programming
… and enhancing coverage and distribution of current applications
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
32
Profiling
ProfilingA definition of profiles that characterize a software service, e.g. software modules, and such profiles that characterize environments, i.e. platforms on which services can be offered, e.g. sensor nodesA definition of profile matching rules defining how these platforms can be reconfigured with these services (in the sense of loadingnew software)
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
33
Profiling – Behavior
a) The robot drives to the position where reconfiguration is necessary
b) It collects information about the environment, builds the context and explores its neighborhood
c) All sensor motes, which have received the exploration message, send their current profiles that contain information about the hardware and software of the node
d) The robot uses the information gathered in steps b) and c) to assign the roles of the sensor motes, optimized for the current goal
e) The robot re-programs selected sensor motes over the air
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
34
Profiling – Algorithm
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
35
Profile Descriptions
node {properties:address = 1;board = mica2;sensors = mts310;appl = LightMeasure;
}
application {properties:name = TempMeasure;modules = TempSensorM,
CalcM;requirements:board = mica2;
}
module {properties:name = TempSensorM;
requirements: sensor = mts310;
}
module {properties:name = TempSensorM;
requirements:sensor = mts101;
}
module {properties:name = CalcM;
}
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
36
Simulation in OMNeT++
Analysis of alternate configurations and large-scale environments
Routing, media accessMobility models and strategiesCoordination, task allocation
Important measuresEnergy efficiency (lifetime)PerformanceAvailability
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
37
Profiling – Conclusions
Location-dependent reconfiguration realized in a prototype
Resource-aware data format necessary and possible
Sensor motes do not have to provide functionality but to store an ID (profile) describing their current HW/SW configuration
Profile matching, code generation, and other resource-intensive tasks are provided by the mobile robot systems (behaving as local servers)
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
38
Outline
Autonomic Networking GroupIntroductionResearch projects
Selected research issuesBio-inspired networking
Feedback-loopCongestion control
ProfilingRobot-assisted WSN
Data communication in WSN(semi-)reliable and authenticated communication in WSN
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
39
Mobile Ad Hoc Networks
CharacteristicsStrong resource limitations (storage, processing power)Error-prone and unreliable communication pathsQuick changes of the communications paths due to mobility
Even stronger limitations in wireless sensor networks (WSN) and WPANs
Necessary enhancementsReliable and semi-reliable communication servicesMechanisms to ensure data integrity and message authentication
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
40
(Semi-)reliable Communications
Reliable communicationTypically based on ACK/NAK schemesUnacknowledged messages are retransmittedDifferent schemes to optimized this process have been studied
Stop-and-wait, sliding-windowTCP (RTO backoff, fast retransmit, …)
Semi-reliable communicationLost messages can be identifiedNo retransmission is initiated by the protocol
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
41
Message Authentication
Required functionalityDetection whether a message was altered during transmissionIdentification of the sender of the message
Available algorithmsModification Detection Code (MDC)
Message Digest 5 (MD5)Secure Hash Algorithm 1 (SHA-1)
Message Authentication Code (MAC)MACs constructed from MDCsReasons for constructing MACs from MDCs
– Cryptographic hash functions generally execute faster than symmetric block ciphers
– There are no export restrictions to cryptographic hash functionsBasic idea: “mix” a secret key K with the input and compute an MDC
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
42
Research Goals
Different mechanisms at different layers or achieved at different locations
Data integrityMessage authenticationReliability(Encryption)
Keeping in mind the resource limitations in WSN/WPAN
RAC
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
43
RAC – Objectives
RAC: (semi-)reliable authenticated communication
Authentication(with key)
Data Integrity(without key)
Sem
i-rel
iabl
e tra
nsm
issi
on(w
ithou
t ret
rans
mis
sion
s)
Rel
iabl
e tra
nsm
issi
on(w
ith re
trans
mis
sion
s)
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
44
RAC – Requirements
Scalabilitythe overhead due to the algorithm should be negligible (message count, message size, memory and processing requirements)
Flexibilitythe optional selection of needed functionality such as reliability vs. semi-reliability and data integrity check vs. full message authentication
Configurabilitythe option to adapt the parameters to the capabilities of the particular entities involved in the communication
Extensibilitythe possibility to implement new functionality such as data encryption to provide confidentiality
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
45
RAC – Working Principle
The sender computes a MDC for every message andStored in a local databaseAppended to the message that is actually transmitted
The receiver used the MDC toVerifies the message integrity / authenticationAcknowledge received messages
Appl.m m
MD5
h+mh+m
Appl.m m
MD5
h hh‘+hhh
A BtACKtRET
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
46
Algorithm Details – Sender
Message sending
for each message mto be sent
docalculate hm=h([K],m)get current time tmstore (tm,hm,[m]) in
databasetransmit (m,hm)
done
Periodically check for lost messages
get current time tcfor each (tm,hm,[m]) in
databasedo
if(tm+tRET>tc)do
retransmit m ornotice lost message
donedone
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
47
Algorithm Details – Receiver
Message receiving
for each receivedmessage (m,hm)
doverify hm -> check
data integrity,messageauthentication
get current time tmstore (tm,hm) in
databasedone
Periodically send acknowledgments
get current time tcfor each (tm,hm) in
databasedo
if(tm+tACK>tc)do
acknowledge allhm in database
donedone
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
48
RAC – Helper Functionality
Modification detection codeMD5, others can be employed as well
Key managementStatic, “ad hoc PKI” demanded for dynamic key management
Flow controlImplicitly defined by tRET and tACK, other groups study TCP-like flow control for ad hoc networks
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
49
Evaluation Model
ParameterstRET – timeout for retransmission or notification of the applicationtACK – timeout for acknowledgmentsλs – data rateρs – loss ratioKs - Keymi – ith messageh([Ks],mi) – MDC of the ith message based on key Ks
λ1, ρ1
λ2, ρ2
tRET tACK
A B|mi, h([K1],mi)|
|mr, h([K2],mr)|
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
50
Simulation Results
Analysis of the size of the retransmission buffer(tRET/tACK=10s/5s, λ=1, variable ρ)
0
2
4
6
8
10
12
0 50 100 150 200time [s]
retr
ansm
issi
on b
uffe
r ρ=0.01ρ=0.05ρ=0.1ρ=0.2
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
51
Simulation Results
Analysis of the size of the retransmission buffer(λ=1, ρ=0.1, variable tRET/tACK)
0
5
10
15
20
0 50 100 150 200time [s]
retr
ansm
issi
on b
uffe
r rRET/tACK: 2s/1srRET/tACK: 5s/2srRET/tACK: 10s/5srRET/tACK: 20s/10s
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
52
Simulation Results
Analysis of the size of the retransmission buffer(tRET/tACK=10s/5s, ρ=0.1, variable λ)
0
10
20
30
40
50
0 50 100 150 200time [s]
retr
ansm
issi
on b
uffe
r λ=1λ=2λ=5
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
53
Simulation Results
Analysis of the overall loss ratio as recognized at the sender of the messages(tRET/tACK=10s/5s, λ=1, variable ρ)
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
0 100 200 300 400 500 600time [s]
ratio
of s
ucce
ssfu
l tr
ansm
issi
ons
ρ=0.01ρ=0.05ρ=0.1ρ=0.2
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
54
RAC – Conclusions
BackgroundAd hoc networks suffer from unreliable data pathsSecurity is a must in many application scenarios
RAC is a new communication method featuring(semi-)reliable data communicationData integrity / message authentication
Studied propertiesRAC is scalable even in WSN because it allows a more efficient utilization of available resources and leads to an improved quality of the global systemSimulation results proved the applicability of the proposed algorithm and allow an on-time adaptation of the individual parameters depending on the current characteristics of the communication pathwaysPrimary application is partial reliability, i.e. the application needs to be informed about loss ratio or about specific lost messages
GWU2005-11-08
Falko DresslerUniversity of Erlangen-Nuremberg, Germany
55
Selected Publications[1] F. Dressler and B. Krüger, "Cell biology as a key to computer networking," German Conference on Bioinformatics
2004 (GCB'04), Bielefeld, Germany, Abstract and Poster, October 2004. [2] F. Dressler, "Efficient and Scalable Communication in Autonomous Networking using Bio-inspired Mechanisms - An
Overview," Informatica - An International Journal of Computing and Informatics, vol. 29 (2), pp. 183-188, July 2005. [3] F. Dressler and G. Carle, "HISTORY - High Speed Network Monitoring and Analysis," Proceedings of 24th IEEE
Conference on Computer Communications (IEEE INFOCOM 2005), Miami, FL, USA, March 2005. [4] F. Dressler, B. Krüger, G. Fuchs, and R. German, "Self-Organization in Sensor Networks using Bio-Inspired
Mechanisms," Proceedings of 18th ACM/GI/ITG International Conference on Architecture of Computing Systems -System Aspects in Organic and Pervasive Computing (ARCS'05): Workshop Self-Organization and Emergence, Innsbruck, Austria, March 2005, pp. 139-144.
[5] F. Dressler, "Sensor-Based Localization-Assistance for Mobile Nodes," Proceedings of 4. GI/ITG KuVSFachgespräch Drahtlose Sensornetze, Zurich, Switzerland, March 2005, pp. 102-106.
[6] F. Dressler and G. Fuchs, "Energy-aware Operation and Task Allocation of Autonomous Robots," Proceedings of 5th IEEE International Workshop on Robot Motion and Control (IEEE RoMoCo'05), Dymaczewo, Poland, June 2005, pp. 163-168.
[7] F. Dressler, "Locality Driven Congestion Control in Self-Organizing Wireless Sensor Networks," Proceedings of 3rd International Conference on Pervasive Computing (Pervasive 2005): International Workshop on Software Architectures for Self-Organization, and Software Techniques for Embedded and Pervasive Systems (SASO+STEPS 2005), Munich, Germany, May 2005.
[8] F. Dressler, "Adaptive network monitoring for self-organizing network security mechanisms," Proceedings of IFIP International Conference on Telecommunication Systems, Modeling and Analysis 2005 (ICTSM2005), Dallas, TX, USA, November 2005.
[9] F. Dressler, "Reliable and Semi-reliable Communication with Authentication in Mobile Ad Hoc Networks," Proceedings of 2nd IEEE International Conference on Mobile Ad Hoc and Sensor Systems (IEEE MASS 2005): International Workshop on Wireless and Sensor Networks Security (WSNS'05), Washington, DC, USA, November 2005.
[10] F. Dressler and I. Dietrich, "Simulative Analysis of Adaptive Network Monitoring Methodologies for Attack Detection," Proceedings of IEEE EUROCON 2005, Belgrade, Serbia & Montenegro, November 2005. (accepted for publication)
[11] F. Dressler and H. Chaskar, "Security Architectures for Wired and Wireless Networks: Threats and Countermeasures," 1st IEEE/ACM International Conference on Communication System Software and Middleware (IEEE COMSWARE 2006), New Dehli, India, Tutorial, January 2006.
[12] B. Krüger and F. Dressler, "Molecular Processes as a Basis for Autonomous Networking," IPSI Transactions on Advances Research: Issues in Computer Science and Engineering, vol. 1 (1), pp. 43-50, January 2005.
Self-Organization in Autonomous Sensor/Actuator
Networks
Falko DresslerAutonomic Networking GroupDept. of Computer Sciences
University of Erlangendressler@informatik.uni-erlangen.de
Bio-inspiredNetworkingBio-inspiredNetworking
Recommended