slide 1/21 dcsl: dependable computing systems lab robust communication primitives in sensor networks...
TRANSCRIPT
Slide 1/21DCSL: Dependable Computing Systems Lab
Robust Communication Primitives in Sensor Networks
Saurabh BagchiDependable Computing Systems Lab
School of Electrical and Computer EngineeringPurdue University
Joint work with: Issa Khalil, Gunjan Khanna, Ravish Khosla, Ness Shroff
http://shay.ecn.purdue.edu/~dcsl
Slide 2/21DCSL: Dependable Computing Systems Lab
Sensor Nodes: Nature of the Beast• Miniature platform for:
• Sensing: Integrated sensor board with sensors for temperature, pressure, humidity, etc.
• Computation: Low power Atmel processor with 128 KB programming and 512 KB data memory
• Communication: Low range ISM band transceiver
• Constraints:– Class I: Energy, Bandwidth, Fragility
– Class II: Processor, Memory
SensorBoard
Radio-ProcessorBoard
InterfaceBoard
Slide 3/21DCSL: Dependable Computing Systems Lab
Dependable Sensor Networking
• Dependability is the property of a system to tolerate failures, be it from natural errors or malicious errors, aka security attacks
Dependability
Resilience to natural errors, i.e.,
Reliability
Resilience to malicious errors,
i.e., Security
Why for Sensors?
1. The nodes are failure prone2. The wireless links are failure prone3. Placed in hazardous environments4. Sometimes used for detection of
critical events
Why for Sensors?1. Placed in hostile environments2. Adversaries have huge gains from
compromising sensor network3. Low cost rules out tamper proof
hardware4. Omni-directional wireless links
Slide 4/21DCSL: Dependable Computing Systems Lab
Application Domains
Domains Sample Applications
Military Target tracking, battlefield surveillance
Civilian
Private Healthcare monitoring, Environmental monitoring, Household security
Public Infrastructure monitoring, Water quality monitoring
Slide 5/21DCSL: Dependable Computing Systems Lab
What is data dissemination?
• There are some sources of sensory data– Possibly sources with overlapping sensing regions
• There are some nodes interested in sensory data– Maybe resource constrained nodes themselves
– Can be cluster heads in hierarchical communication
– Alternately, can be a moving data collector
Control center
Cluster heads
Sensor nodes
Slide 6/21DCSL: Dependable Computing Systems Lab
What is Reliable Data Dissemination?
• Challenge: Need to get data from source to destination– Handle any-to-any communication
– Optimized for common communication pattern
• Capt. Edward Murphy said:– “If a sensor node can fail, it will eventually”
– “If a sensor network link can fail, it will eventually”
• Capt. Edward Murphy also said:– “If a sensor node can move, it will eventually”
• Reliable data dissemination is achieving continuous stream of data from source to destination in the face of the above Murphy’s laws
Slide 7/21DCSL: Dependable Computing Systems Lab
Some Current Approaches for Reliable Data Dissemination
• Flooding
S
N1
N2
N3
N4
N5
N6N7
T
• Con:– Many many redundant transmissions leading to inefficient energy usage
TT
Slide 8/21DCSL: Dependable Computing Systems Lab
Some Current Approaches for Reliable Data Dissemination
• Direct Communication with Base Station
Base StationC
• Con:– Need specialized nodes– Lack of scalability
– Restricted communication pattern
Slide 9/21DCSL: Dependable Computing Systems Lab
Our Approach
• Hybrid of Push and Pull– Push: From source towards sink
– Pull: Interested sink nodes query and pull data from relevant sources
• Approach– Use meta data transmissions to reduce redundant transmissions
– Advertise the data prior to sending the data
– Only interested nodes pull data
– Reduces collisions and energy wastage
S
BADV
C
REQDAT
ADV
S: SenderB: Interested nodeC: Disinterested node
Slide 10/21DCSL: Dependable Computing Systems Lab
Shortest Path Minded SPIN (SPMS)
4
61
2
3
5
ADVADVADVADVADV
REQ
DAT
ADVADV
REQ
DAT
REQ
DAT
• Timers – TimeOutADV:
Nodes wait for the data to come to the nearest node before sending REQ
– TimeOutDAT: Nodes wait for the data after sending the REQ packet
Slide 11/21DCSL: Dependable Computing Systems Lab
SPMS Protocol : Failure Scenario• Resilience to Failures
– After a TimeOutADV expires, node sends the request to PRONE through the shortest path
– DATA is received using the same path if there is no failure
– Incase of a failure TimeOutDAT occurs
– Node directly sends the REQ packet to PRONE
– In case PRONE is also not responding then the REQ is sent to SCONE
• Failures tolerated– Intermediate nodes
– Source node
Slide 12/21DCSL: Dependable Computing Systems Lab
SPMS : Failure Scenario
4
61
2
3
5
ADVADVADVADVADV
REQ
TimeOut_ADVTimeOut_DATREQ
DATA
Slide 13/21DCSL: Dependable Computing Systems Lab
Energy and Delay Analysis• Time to get data from source to adjacent destination is
defined as Tround
S
DADVREQ
DAT
Tround = G.n12 + A.Ttx + Tproc + G.ns
2 + R.Ttx + Tproc + G.ns2 + D.Ttx
Tround = G.n12 + (A+R+D).Ttx + 2Tproc + 2G.ns
2
Slide 14/21DCSL: Dependable Computing Systems Lab
Energy and Delay Analysis• In case of K relay nodes between two nodes
2( 1)
failurefree round ADV cDelay K T TOut T
2 2( ) . 2 . ( ) 2failureround ADV DAT tx proc
Delay k j T TOut G ns TOut G nj R D T T
• The ratio of energy between SPIN and SPMS can be given by :
1
1
:. . . .
r
SPIN SPMSm r
E EE E
k f E k E k E
A C
(k-j+1)th relay node (Failed)
k intermediate nodes
S D
(k-j+1)th relay node (Failed)
k intermediate nodes
ESPIN = (A+D+R).E1 + (A+D+R).Er ESPMS = k.A.E1 + k.(D+R).Em+k.(A+D+R).Er
Slide 15/21DCSL: Dependable Computing Systems Lab
Energy and Delay Comparisons: Equation Plots
SPIN uses more energy than SPMS as relay nodes increase.
Delay advantage of SPMS decreases as relay nodes increase.
Slide 16/21DCSL: Dependable Computing Systems Lab
Simulations• SPMS protocol is simulated in ns-2 and compared with SPIN
– We vary the transmission radius and the number of nodes– Crossbow data sheet is used to calculate the power spent in transmission and
receiving packets.
• Experiments are carried out for two topologies – All to All communication : Every node requests data from every other data– Cluster Based Hierarchical Communication: Cluster heads collect the
data and send it to the sink using SPMS
• Experiments for failure free and failure scenarios– Failures are transient and follow exponential inter-arrival times
• Results– Energy saving with and without failure, with mobility, increases with
increasing sensor field size– Delay improvement increases with increasing sensor field size
Slide 17/21DCSL: Dependable Computing Systems Lab
Optimizations for Failure and Mobility
• Failure optimized SPMS– Avoid sending REQ through a suspected failed path
– Inform neighbors of suspected failed path
• Mobility optimized SPMS– Avoid Bellman Ford on entire zone if node moves in
– Incremental computation in a lazy manner
Slide 18/21DCSL: Dependable Computing Systems Lab
Secure Communication Primitive• Different types of attacks
– Control traffic, vs. Data traffic– Message tampering, eavesdropping, and ID spoofing– Nodes may be compromised
• Symmetric key cryptography can be used• Need to manage the keys
– Energy efficient– Latency sensitive
• Capt. Edward Murphy also said:– “Don’t trust thy neighbor”
Slide 19/21DCSL: Dependable Computing Systems Lab
Our Approach: SECOS
• All the above goals are realized in protocol called SECOS
• Base station is fixed, secure, and has no resource constraints
• All other nodes are generic sensor nodes and have all the typical resource constraints
• Guarantee: Compromising any number of nodes in the network does not compromise the session between two legitimate nodes
Slide 20/21DCSL: Dependable Computing Systems Lab
Take Away Lessons
• Communication protocols in sensor networks have to be designed with – Failures in mind
– Node compromise in mind
• Trade-offs exist between latency and energy consumption and customizable protocols that fit different regions of trade-off curve are desirable
• Desirable characteristics of large class of sensor network communication protocols– No privileged nodes
– No node trusted completely
Slide 21/21DCSL: Dependable Computing Systems Lab
Questions Anyone?
Issa Khalil Gunjan Khanna Ness Shroff
• “Fault Tolerant Energy Aware Data Dissemination Protocol in Sensor Network,” Gunjan Khanna, Saurabh Bagchi, Yu-Sung Wu. At IEEE Dependable Systems and Networks Conference (DSN 2004), June 28-July 1, 2004, Florence, Italy.
• “Analysis and Evaluation of SECOS, A Protocol for Energy Efficient and Secure Communication in Sensor Networks,” Issa Khalil, Saurabh Bagchi, Ness Shroff. Submitted to Ad-hoc Networks Journal, September 2004. Available as CERIAS Tech Report from home page.