DATA AGGREGATION IN WIRELESS SENSOR

NETWORKS
A Seminar Report

Submitted by

P.PRAVEEN KUMAR

(09751A0581)

in partial fulfillment of the requirements for the award of the degree

Of

BACHELOR OF TECHNOLOGY

IN

COMPUTER SCIENCE AND ENGINEERING

At

SREENIVASA INSTITUTE OF TECHNOLOGY AND

MANAGEMENT STUDIES, CHITTOOR-517127
DEC-2012

1
 TABLE OF CONTENTS

3 Data Aggregation: An Overview 8

3.1 In-Network Aggregation

3.2 Tree-Based Approach

3.3 Cluster-Based Approach

3.4 Multi-path Approach

S NO. TITLE PAGE NO.

1 Introduction 5
2 Goal & Problem Definition in Data Aggregation 7
4 Query Processing 15

4.1 Query Models

4.2 Query Language in Tiny DB

4.3 Queries and Aggregates

5 Simulation and Experimental Analysis 20

2
6 Security Challenges 25

7 Advantages and Disadvantages 29

8 Future Scope &Applications 30

9 Conclusion 31

10 References 32

ABSTRACT

Sensor networks are collection of sensor nodes which co-operatively send sensed data to
base station. As sensor nodes are battery driven, an efficient utilization of power is essential in
order to use networks for long duration hence it is needed to reduce data traffic inside sensor
networks, reduce amount of data that need to send to base station. The main goal of data
aggregation algorithms is to gather and aggregate data in an energy efficient manner so that
network lifetime is enhanced. Wireless sensor networks (WSN) offer an increasingly Sensor
nodes need less power for processing as compared to transmitting data. It is preferable to do in
network processing inside network and reduce packet size. One such approach is data
aggregation which attractive method of data gathering in distributed system architectures and
dynamic access via wireless connectivity.

Wireless sensor networks have limited computational power and limited memory and battery
power, this leads to increased complexity for application developers and often results in
applications that are closely coupled with network protocols. In this paper, a data aggregation
framework on wireless sensor networks is presented. The framework works as a middleware for
aggregating data measured by a number of nodes within a network.

3
The aim of the proposed work is to compare the performance of TAG in terms of energy
efficiency in comparison with and without data aggregation in wireless sensor networks and to
assess the suitability of the protocol in an environment where resources are limited.

4
 1. INTRODUCTION

Sensor Networks:
The wireless sensor network is ad-hoc network. It consists small light weighted wireless nodes
called sensor nodes, deployed in physical or environmental condition. And it measured physical
parameters such as sound, pressure, temperature, and humidity. These sensor nodes deployed in
large or thousand numbers and collaborate to form an ad hoc network capable of reporting to
data collection sink (base station). Wireless sensor network have various applications like habitat
monitoring, building monitoring, health monitoring, military survivallance and target tracking.
However wireless sensor network is a resource constraint if we talk about energy, computation,
memory and limited communication capabilities. All sensor nodes in the wireless sensor network
are interact with each other or by intermediate sensor nodes.

With advance in technology, sensor networks composed of small and cost effective sensing
devices equipped with wireless radio transceiver for environment monitoring have become
feasible. The key advantage of using these small devices to monitor the environment is that it
does not require infrastructure such as electric mains for power supply and wired lines for
Internet connections to collect data, nor need human interaction while deploying. These sensor
nodes can monitor the environment by collecting information from their surroundings,and work
cooperatively to send the data to a base station, or sink, for analysis.

5
 Figure 1 Architecture of the Sensor network

A sensor nodes that generates data, based on its sensing mechanisms observation and transmit
sensed data packet to the base station (sink). This process basically direct transmission since the
base station may located very far away from sensor nodes needs. More energy to transmit data
over long distances so that a better technique is to have fewer nodes send data to the basestation.
These nodes called aggregator nodes and processes called data aggregation in wireless sensor
network.

6
 2. GOALS AND PROBLEM DEFINITION

GOAL:

The main goal of data aggregation algorithms is to gather and aggregate data in an energy
efficient manner so that network lifetime is enhanced. Wireless sensor networks (WSN) offer an
increasingly attractive method of data gathering in distributed system architectures and dynamic
access via wireless connectivity.

PROBLEM DEFINITION:

Data aggregation protocols aims at eliminating redundant data transmission and thus improve the
lifetime of energy constrained wireless sensor network. In wireless sensor network, data
transmission took place in multi-hop fashion where each node forwards its data to the neighbor
node which is nearer to sink. Since closely placed nodes may sense same data, above approach
cannot be considered as energy efficient. An improvement over the above approach would be
clustering where each node sends data to cluster-head (CH) and then cluster-head perform
aggregation on the received raw data and then send it to sink. Performing ag homogeneous
sensor network cluster-head will soon die out and again re-clustering has to be done which again
cause energy consumption.

7
 3. DATA AGGREGATION:AN OVERVIEW

Data aggregation is a process of aggregating the sensor data using aggregation approaches. The
general data aggregation algorithm works as shown in the below figure. The algorithm uses the
sensor data from the sensor node and then aggregates the data by using some aggregation
algorithms such as centralized approach, LEACH(low energy adaptive clustering
hierarchy),TAG(Tiny Aggregation) etc. This aggregated data is transfer to the sink node by
selecting the efficient path

8
 Fig:General architure of algorithm

3.1 In-Network Aggregation:

In-network aggregation is the global process of gathering and routing information through a
multi-hop network, processing data at intermediate nodes with the objective of reducing resource
consumption (in particular energy), thereby increasing network lifetime.

There are two approaches for in-network aggregation:

1) with size reduction

2) without size reduction.

With size reduction:

In-network aggregation with size reduction refers to the process of combining
and compressing the data packets received by a node from its neighbors in order to reduce the
packet length to be transmitted or forwarded towards sink

Without size reduction:

In-network aggregation without size reduction refers to the process merging data packets
received from different neighbors in to a single data packet but without processing the value of
data.

9
3.2 Tree Based Approach:

The tree based approach is defining aggregation from constructing an aggregation tree. The form
of tree is minimum spanning tree, sink node consider as a root and source node consider as a
leaves. Information flowing of data start from leaves node up to root means sink(base
station).Disadvantage of this approach, as we know like wireless sensor network are not free
from failure .in case of data packet loss at any level of tree, the data will be lost not only for
single level but for whole related sub tree as well.This approach is suitable for designing optimal
aggregation techniques data centric protocol know as Tiny aggregation (TAG) approach.

Fig:Tree-based aggregation in wireless sensor networks

10
The working of TAG is depending on two phases:

1) distributed phase

2) collection phase

Distributed Phase:

In distributedphase, in which aggregate queries are pushed down into the network.

Collection Phase:
A collectionphase, where the aggregate values are continually routed up from children to parents.
Recall that our query semantics partition time into epochs of duration ,and that we must produce
a single aggregate value (when not grouping) that combines the readings of all devices in the
network during that epoch.

11
3.3 Cluster-Based Approach:

In energy-constrained sensor networks of large size, it is inefficient for sensors to transmit the
data directly to the sink In such scenarios, Cluster based approach is hierarchical approach. In
cluster-based approach, whole network is divided in to several clusters. Each cluster has a
cluster-head which is selected among cluster members. Cluster-heads do the role of aggregator
which aggregate data received from cluster members locally and then transmit the result to base
station (sink). Recently, several cluster-based network organization and data-aggregation
protocols have been proposed for the wireless sensor network.

Fig:Cluster Based sensor networks.Arrow indicates wireless communication links

12
The cluster heads can communicate with the sink directly via long range transmissions or multi
hopping through other cluster heads.

K. Dasgupta in proposed a maximum lifetimedata aggregation (MLDA) algorithm which finds

datagathering schedule provided location of sensors node andbase-station, data packet size, and
energy of each sensornode. A data gathering schedule specifies how data packetare collected
from sensors node and transmitted to basestation for each round. A schedule can be thought of as
acollection of aggregation trees. They proposedheuristic-greedy clustering-based MLDA based
on MLDAalgorithm. In this they partitioned the network in to clusterand referred each cluster as
super-sensor. They thencompute maximum lifetime schedule for the super-sensorsand then use
this schedule to construct aggregation treesfor the sensors. W. Choi et present a two-
phaseclustering (TPC) scheme. Phase I of this scheme createsclusters with a cluster-head and
each node within thatcluster form a direct connects with cluster-head. Phase I thecluster-head
rotation is localized and is done based on the remaining energy level of the sensor nodes which
minimizetime variance of sensors and this lead to energy savingfrom unnecessary cluster-head
rotation. In phase II, eachnode within the cluster searches for a neighbor closer thancluster-head
which is called data relay point and setup up adata relay link.

Now the sensor nodes within a clustereither use direct link or data relay link to send their data
tocluster head which is an energy efficient scheme. The datarelay point aggregates data at
forwarding time to another data relay point or cluster head. In case of high networkdensity, TPC
phase II will setup unnecessary data relaylink between neighbors as closely deployed sensor
willsense same data and this lead to a waste of energy.

13
3.4 Multi-Path Approach:

The drawback of tree based approach is the limited robustness of the system. To overcome this
drawback, a new approach was proposed by many researchers .in which sending partially
aggregated data to single parent node in aggregation tree, a node could send data over multiple
paths. In which each and every node can send data packets to its possibly multiple neighbors.
Hencedata packet flow from source node to the sink node along multiple path, lot of intermediate
node between source node to sink node so aggregation done in every intermediate node.

Using this approach we will make the system robust but some extra overhead. The example of
this approach like ring topology, where network is divided in to concentric circle with defining
level levels according to hop distance from sink.propose a new strategy have both issues : energy
efficiency and robustness. In which single path to connect each node to the base station it is
energy saving but high risk of link failure. But on the other head multipath approach would
require more nodes to participate with consequent waste of energy. Authors present a clever use
of multi-path only when there is loss of packet which is implemented by smart caching of data at
sensor nodes. Authors also argue that in many practical situation data may be gathered only from
a particular region, so they use a different approach that relies on a spanning tree and provides
alternative paths only when a malfunctioning is detected. Algorithm adopts a tree-based
approach for forwarding packets through the network. In the ideal situation when no failures
occur, this is certainly the best choice, as the minimum number of nodes is engaged in the
transmission phase. In the presence of link or node failures, the algorithm will discover
alternative paths, so as ensure the delivery of as many packets as possible within the time
constraints.

The problem with this approach is that it may cause the arising of hot spots and nodes along
preferred paths will consume their energy resources quickly, possibly causing disconnection in
the network.

14
 4. QUERY PROCESSING

4.1 Query Models:

COUGAR approach proposes a query layer tosupport aggregate queries. With the interface
provided, the clients can issue queries without knowing how the results are generated, processed
and returned by the sensor network to them. The query layer processes declarative queries and
generate a cost effective query plan. They follow a database approach to design a query interface
for sensor networks. The view of cost is different for sensor networks.

The major factor under consideration is the communication cost, involving the cost of routing the
queries and aggregating data over the sensor networks. TAG also proposes a query model for
supporting aggregate queries. TAG and COUGAR are tightly coupled with the underlying
aggregation schemes. Proposes a Query Agent that provides application independent query
interface and an API support to map the user specified queries to lower level semantics
corresponding to underlying routing and aggregating protocols. It supports different
communication models - anycast, unicast, multicast and broadcast. Query agent will support a
wide variety of routing and aggregation protocols selecting the best combination based on the
type of the query.

15
4.2 Query Language in Tiny DB:

TinyDBs query language is based on SQL, and we will refer to it as TinySQL. Query Language
in TinySQL supports selection, projection, determining sampling rate, group aggregation, user
defined aggregation, event trigger, lifetime query, setting storing point and simple join.

The query language of Antelope is called AQL and isused both to build and to query databases.
The typical way of using Antelope in a sensor device is first to define a database consisting of
one or more relations. Data can be modeled using the well-founded design principles for
relational databases, such as normalizing the data and planning for what kind of queries that the
system should be able to handle efficiently.

Table 1. Antelope database operations

Operation Purpose
INSERT Insert a tuple into a relation.
REMOVE Remove tuples matching a condition.
SELECT Select tuples and project attributes.
JOIN Join two relations on a condition.
CREATE RELATION Create an empty relation.
REMOVE RELATION Remove a relation and all its associated indexes.
CREATE ATTRIBUTE Add an attribute to a relation.
CREATE INDEX Create an attribute index.
REMOVE INDEX Remove an attribute index.

16
1CREATE RELATION sensor;

2CREATE ATTRIBUTE id DOMAIN INT IN sensor;

3 CREATE ATTRIBUTE name DOMAIN STRING(20) IN sensor;
4 CREATE ATTRIBUTE position DOMAIN LONG IN sensor;

5 CREATE INDEX sensor.id TYPE INLINE;

6 CREATE INDEX sensor.positionTYPE MAXHEAP;

Example 4.1: A database in Antelope is created by issuing a series of catalog operations. We first
create the relation, then its attributes, and lastly its indexes.

AQL, Users define databases by using a set of catalog op-erations provided by the language. The
database can then be queried by using a set of relational operations. Several of these operations
share syntactic elements with SQL, but there is also a considerable difference caused by the
different goals of the languages. SQL targets a broad range of database systems, including high-
end systems with several orders of magnitude larger resources than sensor devices.

AQL, by contrast, is designed for systems with modest hardware resources. Hence, complex
SQL functionality such as procedural extensions, triggers, and transactions are first verifies that
all its attribute values pertain to the domain of the corresponding attribute. The abstract
representation in AQL is thereafter transformed into its physical representation, which is a
compact byte-array record of the attribute values. If the attribute is indexed, the Antelope calls
the abstract function for index insertion, which forwards the call to the actual index component
chosen for the attribute. In the final step, the tuple is passed to the storage abstraction,
which writes it to persistent storage.

17
4.3 Queries and Aggregates:

The probable queries for the sensor networks can be categorized into:

1) Simple Queries:
These are non aggregate queries.
Eg. "SELECT temperature FROM sensor WHERE node = z". These are generally mapped into
broadcast or point to point queries.

2) Complex Queries:
They may contain sub queries.
Eg. "SELECT temperature FROM sensor WHERE room =(SELECT room WHERE floor =
3)".

3) Event Driven Queries:

These are the continuous query that returns the values periodically at specified time intervals.
Eg: SELECT AVG (temperature) FROM sensor where node =z" .The Grammar of TinySQL
query language is as follows:

SELECT select-list [FROM sensors] WHERE predicate 294 [GROUP BY gb-list] [TRIGGER
ACTION command-name[(param)]] [EPOCH DURATION time]

Where, selectlist is the attribute list of the unlimited virtual relational table, which can include
an aggregation function. Predicate is the query condition.gblist is an attributes list.
commandname is a trigger operation. Param is the parameters of trigger. Time is the value of
time. TRIGGERACTION is the subordinate clause which defines the trigger. It determines the
operations executed when WHERE clause is satisfied. EPOCH DURATION defines the query
cycle. The meaning of the other clauses is the same as SQL.

18
Following is an example of a TinyDB query.

SELECT nodeid, AVG(light), AVG(temp) FROM sensors WHERE AVG(light)=100 GROUP BY

nodeidEPOCH DURATION 5min

The meaning of the query is detecting nodeid per five minutes in which the average light is equal
to 100 and returning the nodeid and its average light and temperature. Currently, the functions of
TinyDB are very limited. Some functions supported by SQL are not supported by TinyDB.

19
 5. SIMULATION AND EXPERIMENTALANALYSIS

Simulation Tools:

We have plenty of simulation tools or simulators for simulating wireless networks. The
simulators which are most popular are TOSSIM, NS-2, OPNET, OMNet++,J Sim, GlomoSim,
and Qualnet and so on. TOSSIM is a discrete event simulator for TinyOS (TinyOS is a popular
sensor network operating system) sensor networks. Instead of compiling a TinyOS application
for a mote, users can compile it into the TOSSIM [20] framework, which runs on a PC. This
allows users to debug, test, and analyze algorithms in a controlled and repeatable environment.
As TOSSIM runs on a PC, users can examine their TinyOS code using debuggers and other
development tools. TOSSIMs primary goal is to provide a high fidelity simulation of TinyOS
applications. For this reason, it focuses on simulating TinyOS and its execution, rather than
simulating real word.

Simulation run:
This simulation is run for the following with aggregation and clustering Query-1.

20
QUERY-1: SELECT AVG (light) FROM SENSORS GROUP BY NODEID % 2 SAMPLE
PERIOD 2048

Fig: Result window for with aggregation and clustering

21
QUERY-2: SELECT MAX (temp), AVG (light) FROM SENSORS SAMPLE PERIOD 2048

Fig: Result window for with aggregation and without clustering

22
QUERY-3: SELECT temp, light FROM SENSORS SAMPLE PERIOD 2048

Simulation results and comparison:

With aggregation query:

SELECT MAX (temp), AVG (light) FROM SENSORS SAMPLE PERIOD 2048

Without aggregation query:

23
 SELECT light FROM sensors SAMPLE PERIOD 2048

With aggregation and with clustering query:

SELECT AVG (light) FROM SENSORS GROUP BY NODEID %2 SAMPLE PERIOD

2048.

24
 6. SECURITY CHALLENGES

Data acquisition systems for sensor networks can be classified into two broad categories on the
basis of the data collection methodology employed for the application:

Query-based systems:
In query-based systems, the base station (the data sink) broadcasts a query to the network and the
nodes respond with the relevant information. Messages from individual nodes are potentially
aggregated enroute to the base station. Finally, the base station computes one or more aggregate
values based on the messages it has received.

In some applications, queries may be persistent in nature resulting in a continuous stream of data
being relayed to the data sink from the nodes in the network. For such applications, the query
broadcast by the base station specifies a period (referred to as an epoch); nodes in the network
send their readings to the base station after each epoch.

Event-based systems:
In event-based applications, such as perimeter surveillance and biological hazard detection,
nodes send a message to the base station station only when the target event occurs in the area of
interest. If multiple reports being relayed correspond to the same event, they can be combined by
an intermediate node on the route to the base station.

Data acquisition systems can also be categorized based on how sensor data is aggregated. In
single-aggregator approaches, aggregation is performed only at the data sink

In contrast, hierarchical aggregation approaches make use of in-network

aggregation.Hierarchical aggregation schemes can be further classified into tree-based schemes
and ring-based schemes on the basis of the topology into which nodes are organized.

25
As discussed in the introduction, most existing data management and acquisition systems for
sensor networks are vulnerable to security attacks launched by malicious parties. Sensor nodes
are often deployed in unattended environments, so they are vulnerable to physical tampering.
Since current sensor nodes lack hardware support for tamper resistance, it is relatively easy for
an adversary to compromise a node without being detected.

The adversary can obtain confidential information (e.g., cryptographic keys) from the
compromised sensor and reprogram it with malicious code. Moreover, the attacker can replicate
the compromised node and deploy the replicas at various strategic locations in the network

A compromised node can be used to launch a variety of security attacks. These attacks include
jamming at the physical or link layer as well as other resource consumption attacks at higher
layers of the network software. Compromised nodes can also be used to disrupt routing protocols
and topology maintenance protocols that are critical to the operation of the network .

In this topic we focus on attacks that targetthe data acquisition protocol used by the application.
Specifically, we discuss attacks in which the compromised nodes send malicious data in response
to a query (or send false event reports in event-based systems). By using a few compromised
nodes to render suspect the data collected at the sink, an adversary can effectively compromise
the integrity and trustworthiness of the entire sensor network.
\

In event-based systems, compromised nodes can be used to send false event reports to the base
station with goal of raising false alarms and depleting the energy resources of the nodes in the
network. We refer to this attack as the false data injection attack. we discuss an approach for
filtering the false data injected by malicious nodes
.
Similarly, in query-based systems, compromised nodes can be used to inject false data into the
network with the goal of introducing a large error in the aggregate value computed at the data
sink. The aggregate computed by the sink is erroneous in the sense that it differs from the true

26
value that would have been computed if there were no false data values included in the
computation. Unlike event-based systems, however, in aggregation systems the effectiveness of a
false data injection attack depends on both the aggregate being computed, e.g., MAX or
MEDIAN, and whether sensor data is aggregated enroute to the data sink or only at the data sink,
and thus the techniques used for preventing this attack differ from the techniques used in event-
based systems.

In an aggregation system, a compromised node M can corrupt the aggregate value computed at
the sink in four ways. First, M can simply drop aggregation messages that it is supposed to relay
towards the sink. This has the effect of omitting a large fraction of sensor readings being
aggregated. Second, M can alter a message that it is relaying to the data sink. Third, M can
falsify its own sensor reading with the goal of influencing the aggregate value. Fourth, in
systems that use in-network aggregation, M can falsify the sub-aggregate which it is supposed to
compute based on the messages received from its child nodes.

The first attack in which a compromised node intentionally drops aggregation messages can
substantially deviate the final estimate of the aggregate if tree-based aggregation algorithms are
used. The deviation will be large if the compromised node is located near the root of the
aggregation hierarchy because a large fraction of sensor readings will be omitted from being
aggregated. Countermeasures against this attack include the useof multi-path routing and ring-
based topologies as well as the use of probabilistic techniques in the formation of aggregation
hierarchies

To prevent the second attack in which a compromised node alters a message being relayed to the
sink, it is necessary for each message to include a message authentication code (MAC) generated
using a key shared exclusively between the originating node and the sink. This MAC enables the
sink to check the integrity of a message, and filter out messages that have been altered. Hence,
the effect of altering a message is no different from dropping it, and countermeasures such as
multi-path routing are needed to mitigate the effect of this attack.

27
We refer to the third attack in which a sensor intentionally falsifies its own readingas the falsified
local value attack. This attack is similar to the behavior of nodes with faulty sensors, and also to
the false data injection attack in event-based systems. Potential countermeasures to this attack
include approaches used for fault tolerance such as majority voting and reputation-based
frameworks. we discuss how aggregation schemes can be designed to be resilient to this attack.

The three attacks discussed above apply to both single-aggregator and hierarchical aggregation
systems, whereas the fourth attack applies only to hierarchical aggregation systems. This attack
in which a node falsifies the aggregate value it is relaying to its parent(s) in the hierarchy is much
more difficult to address. We refer to this attack as the falsified sub-aggregate attack. we discuss
two approaches that have been proposed for resilient hierarchical aggregation that include
countermeasures for this attack.

28
 8. ADVANTAGES AND DISADVANTAGES

Advantages:

1)With the help of data aggregation process we can enhance the robustness and accuracy of
information which is obtained by entire network, certain redundancy exists in the data collected
from sensor nodes thus data fusion processing is needed to reduce the redundant
information.

2) Another advantage is those reduces the traffic load and conserve energy of the sensors.

Disadvantages:

1) Thecluster head means data aggregator nodes send fuse these data to the base station .this
cluster head or aggregator node may be attacked by malicious attacker.

2) If a cluster head is compromised, then the base station (sink) cannot be ensure the correctness
of the aggregate data that has been send to it.

3) Another drawback is existing systems are severalcopies of the aggregate result may be sent to
the base station (sink) by uncompromised nodes .It increase the power consumed at these nodes.

29
 9. FUTURE SCOPE

Future work will focuses on the using new different routing algorithms for routing the data from
the source to the sink.

Our approach should confront with the difficulties of topology construction, data routing, loss
tolerance by including several optimization techniques that further decrease message costs
andimprove tolerance to failure and loss.

In addition to implementing these techniques, we need to rethink some of these techniques to

present more efficiency to network changes and external factors which could affect our approach
such as node mobility, obstacles and other issues.

In addition as future work, we could also extend our simulator to incorporate a 3D tree
construction technique.

30
 10. CONCLUSION

In this work we have studied the two most important parts ofdata communication in sensor
networks- query processing, data aggregation and realized how communication in sensor
networks is different from other wireless networks. Wireless sensor networks are energy
constrained network. Since most of the energy consumed for transmitting and receiving data, the
process of data aggregation becomes an important issue and optimization is needed. Efficient
data aggregations not only provide energy conservation but also remove redundancy data
and hence provide useful data only.

The simulation result shows that when the data from source node is send to sink through
neighbors nodes in a multihopfashion by reducing transmission and receiving power, the energy
consumption is low as compared to that of sending data directly to sink that is aggregation
reduces the data transmission then the without aggregation. We have showed how aggregate
queries are efficiently executed in wireless sensor networks.

31
 11. REFERENCES

1. S. Lindsey and C. Raghavendra, PEGASIS: Powerefficient gathering in sensor

information systems, in 2010 IEEE International Conference on Computational
Intelligence and Computing Research Proceedings of IEEE AerospaceConference, vol. 3,
Mar. 2002, pp. 11251130.

2. M. Lee, and S. Lee, Data Dissemination for Wireless Sensor Networks, in Proceedings
of the 10 th IEEEInternational Symposium on Object and Component-Oriented Real-
Time Distributed Computing(ISORC07).

3. E. Fasolo, M. Rossi, J. Widmer, and M. Zorzi, In-Network Aggregation Techniques for

Wireless Sensor Networks: ASurvey, IEEE Wireless communication 2007.

4. C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, and F. Silva, Directed for

Wireless Sensor Networking, IEEE/ACMTransactions on Networking.

5. H. Cam, S. Ozdemir, P. Nair, and D.Muthuavinashiappan, ESPDA: Energy-Efficient and

Secure Pattern-based Data Aggregation for Wireless Sensor Networks,in proceedings of
IEEE Sensor- TheSecond IEEE Conference on Sensors, Toronto,Canada.

6. ChalermekIntanagonwiwat, Ramesh Govindan, and Deborah Estrin, Directed diffusion:

a scalable and robust communication paradigm for sensor networks,(MobiCom 2000)

7. K. Dasgupta, K. Kalpakis, and P. Namjoshi, An Efficient Clustering-based Heuristic for

Data Gathering and Aggregation in Sensor Networks, IEEE 2003.

8. Funda Ergun, SampathKannan, S. Ravi Kumar, RonittRubinfeld, and Mahesh

Viswanathan. Spotcheckers.JCSS, 60:717751, 2000.

32
9. Daniel J. Abadi, Wolfgang Lindner, Samuel Madden, and Jorg Schuler. An integration
framework for sensor networks and data stream management systems. In Mario A.
Nascimento, M. Tamer O zsu, Donald Kossmann, Renee J. Miller, Jose A. Blakeley,
and K. Bernhard Schiefer, editors, VLDB, pages 13611364. Morgan Kaufmann, 2004.

10. H.H. Yen, C.L. Lin: Integrated channel assignment and data aggregation routing problem
in wireless sensor networks,IET Communications, 2009.

11. Considine, F. Li, G. Kollios, and J. Byers. Approximate aggregation techniques for sensor
databases.In Proc. of IEEE Intl Conf. on Data Engineering (ICDE), 2004.

12. E. Fasolo, M. Rossi, J. Widmer, and M. Zorzi, In-Network Aggregation Techniques for
Wireless Sensor Networks: A Survey, IEEE Wireless communication,2007.

13. M. Lee and V.W.S. Wong, An Energy-aware Spanning Tree Algorithm for Data
Aggregation in Wireless Sensor Networks, IEEE PacRrim 2005, Victoria, BC, Canada,
Aug. 2005.

14. M. Ding, X. Cheng, and G. Xue, Aggregation tree construction in sensor networks, in
Proc. of IEEE VTC03, Vol. 4, Orlando, FL, Oct. 2003.

15. The Design Space of Wireless Sensor Networks by Kay Romer and
FriedemannMatternhttp://www.vs.inf.ethz.ch/publ/papers/wsndesignspace.pdf

33