Title: Dependable Computing Systems Lab Middleware for Robust Sensor Networks
1Dependable Computing Systems LabMiddleware for
Robust Sensor Networks
- School of Electrical Computer Engineering
- Purdue University
Members Faculty Saurabh Bagchi Students Nipoon
Malhotra Gunjan Khanna Yu-Sung Wu Issa
Khalil Yen-Shiang Shue Jin-Yi Wang Bingrui
Foo Blake Matheny
URL http//shay.ecn.purdue.edu/dcsl/
2Data Dissemination in Sensor Networks
- Large part of a sensor networks role is sensor
data gathering and dissemination - The data is often critical and has soft real-time
requirements - The paths on which the data traverses are often
unreliable nodes and links may fail transiently
or permanently - Motivation of robust data dissemination
- Goals of data dissemination protocols
- Minimize energy drain
- Distribute data among the nodes quickly
- Reduce redundant data transmissions
- Tolerate node and link failures
3Existing Approaches to Data Dissemination
- TTDD Sources are the sensor nodes and sinks are
mobile nodes interested in sensor data - Sets up a grid structure and proactively
determines routing from data source to sink - At runtime, when sink needs data it locates a
near dissemination point which uses
pre-computed route from source to sink - Drawbacks Cost of setting up entire routing
grid. - LEACH - Clustering of nodes for data forwarding
to base station - Clusters formed and cluster heads chosen
- Data forwarded to cluster head in TDMA manner,
which is responsible for sending data to sink - Drawbacks Data exchange on pre-determined
schedule, all nodes need to be able to
communicate directly with base station
- PEGASIS - Single node responsible for sending
entire data to the sink - Aggregate data from all cluster heads at single
node - Drawbacks Uneven drain of energy, higher delay
- SPIN - Exchange of meta data prior to actual
data exchange, mix of push-pull - Advertisement and request with meta-data before
data sent only to interested nodes - Drawbacks Single hop communication leading to
high energy expenditure and contention delay
ADV
SPIN
B
REQ
DAT
S
S Sender B Interested node C Disinterested node
ADV
C
4Our Approach SPMS
- Single hop communication is inefficient in energy
consumption and delay - Forming efficient clusters and reconfiguring them
in the face of motion are difficult - Our protocol Shortest Path Minded SPIN (SPMS)
- Use meta data to avoid redundant data
transmissions - Incorporate multi-hop communication to use the
available multiple transmit power levels - Reduce energy and latency (due to MAC contention)
by using smaller power.
SPMS
Pmax(ADV)
P1(REQ)
B
S
I1
I2
P1(DAT)
P1 lt Pmax
5Results Energy Dissipation Delay
Failure Free Cases
Failure Case
6SensorNMR A Data Fault Tolerant Approach In
Sensor Networks
- Motivation
- Protect integrity of sensed data reaching the
sink node - Error correction is wasteful in bandwidth, energy
and requires complicated hardware and/or software - Using N-Modular Redundancy (NMR) utilizes the
redundancy inherent in the broadcast nature of
wireless sensor network protocols - Our Approach
- Vote at the sink node on data from multiple
redundant paths to achieve data fault tolerance - Competing idea use a hybrid-ARQ scheme (H-ARQ)
- H-ARQ is a method that combines FEC with
automatic repeat request (ARQ)
7Problem With H-ARQ
- Energy consumption of H-ARQ scheme is much worse
than NMR schemes
- Energy consumption for some H-ARQ schemes and
broadcast TMR against different bit error rates
for different grid (network)sizes
8The Directed Gossip Protocol
- NMR is applicable to current sensor network
protocols - Broadcast protocol There are more redundant
packets than necessary for voting at the sink, so
wasteful power is wasted in transmitting them - Directed diffusion protocol Although sensors
have a sense of direction in this protocol, there
are still more packets than needed for voting at
the sink - Gossip Packets have no directional sense and so
will take a long time to converge to the
destination - Solutions Directed Gossip Protocol!
- Combines the directional sense of directed
diffusion with advantage of low energy cost of
gossip protocol
Sink
Source
9Results Reliability, Convergence Time Energy
Usage
(C) Energy Usage
(B) Convergence Time
CWSA Center for Wireless Systems and Applications
10Scalable Energy Efficient Crypto on Sensors
(SECOS)
- Sensors deployed in hostile environments where
communication susceptible to eavesdropping - Solution Protect message communication using
symmetric key cryptography - Challenge How to manage keys in a manner that is
- Scalable
- Secure
- Resource Aware
- Current Approaches
- Every sensor node has the keys of all other nodes
(poor security, poor scalability, high storage
requirements) - Using the sink to hand the keys to parties
wanting to communicate (not scalable,
energy-expensive, no key refreshment)
11Our Approach SECOS
- Use a structured hierarchy of nodes divided into
control groups - Control node chosen randomly by base station
- Control node changed frequently
- Control node provides key management for sensor
nodes in group - Communication within control group is optimized
- Communication across control group involves
secure session between control nodes - Keys refreshed frequently
M
M
Base station
K1,i
K2,i
Ci
K3,i
Control node
C2
C3
C1
Sensor node
Si,j
S3,1
S3,1
S2,n
S1,n
S2,1
S1,1