Node Legitimacy Based False Data Filtering Scheme in Wireless Sensor Networks

Int. J. Advanced Networking and Applications
Volume: 6 Issue: 5 Pages: 2435-2439 (2015) ISSN: 0975-0290
2435
Node Legitimacy Based False Data Filtering
Scheme in Wireless Sensor Networks
YaFang Liu
College of Environmental Science & Engineering, Ocean University of China, China
Email: lliuyafang@126.com
HaiPeng Qu
College of Environmental Science & Engineering, Ocean University of China, China
Email: haipeng.qu@gmail.com
----------------------------------------------------------------------ABSTRACT-----------------------------------------------------------
False data injection attack is a serious threat to wireless sensor network. In this paper, a node legitimacy based
false data filtering scheme (NLFS) is proposed. NLFS verifies not only message authentication codes (MACs)
contains in reports, but also the legitimacy of nodes that endorse the report. The verification guarantees that
compromised nodes from different geographical areas cannot collude to inject false data, which makes NLFS has a
high tolerance of compromised nodes. In addition, NLFA only utilizes the relationships between node IDs to verify
the legitimacy of nodes without other software or hardware overhead. Simulation results show that NLFS can filter
95% false reports within three hops and is resilience to an increasing number of compromised nodes.
Keywords - compromised node, false data injection, node legitimacy, wireless sensor networks;
-------------------------------------------------------------------------------------------------------------------------------------------------
Date of Submission: January 17, 2015 Date of Acceptance: February 15, 2015
-------------------------------------------------------------------------------------------------------------------------------------------------
1 Introduction
Wireless sensor networks (WSNs) consist of a large
number of sensor nodes with limited resources. Sensor
nodes are usually deployed at unattended or hostile
environments. Therefore, they have a high risk of being
captured and compromised. An Adversary can access all
keying materials stored in compromised nodes, and utilize
these compromised nodes to send bogus reports to the sink
causing false alarm, wrong decision, as well as energy
waste in forwarding nodes.
Some false data filtering schemes [1-13] have been
proposed recently. According to the encrypting
mechanisms, the related work falls into two categories:
symmetric key based schemes and asymmetric key based
schemes. Although asymmetric key based schemes [1-3]
offer superior security, most of them are not practical, since
these schemes are computation intensive and sensor nodes
have limited computing power and restricted memory space.
Symmetric key based schemes [3-13] share a general
en-route filtering framework. In this framework, nodes first
establish key-sharing relationships. When an event happens,
at least t nodes collaboratively generate a report. Each node
attaches its MAC to the report as an endorsement. Here t is
a security threshold. Forwarding nodes utilize the key
sharing relationships to verify the correctness of the MACs
in the report, and reports that contain wrong MACs will be
dropped by the forwarding nodes or sink. However, most of
these schemes only consider the correctness of MACs,
which makes that any t compromised nodes even from
different geographical areas can collude to inject false
reports that cannot be detected. Existing schemes [7, 8, 12]
that consider the legitimacy of nodes require that every
participating node has self-positioning capability. The
scheme proposed in [13] utilizes relative positions of
sensors to verify the legitimacy of nodes, but it requires fix
path between the Sink and each cluster header, which is not
practical due to frequent routing changes. In addition, it is
probabilistic since it cannot guarantee that every false
report will be filtered during the en-route filtering phase.
In this paper, we propose node legitimacy based false
data filtering scheme (NLFS). The objectives of NLFS are
summarized as follow:
First, NLFS initializes serial ID numbers for nodes in
the same cluster; forwarding nodes only utilize the
relationships between IDs to verify the legitimacy of nodes
without other software or hardware overhead.
Second, NLFS can offer stronger filtering capacity and
drop false data within certain number of hops.
Third, NLFS can defend against collaborative false data
injection attack launched by compromised nodes from
different geographical areas and has a high tolerance of
compromised nodes.
2 System model and threat model
2.1 System model
We consider a sensor network composed of a large number
of sensor nodes and these nodes are organized into clusters
after deployment [14]. We assumes that the sensor nodes
are deployed in high density, so that each cluster can
contain at least t nodes and one of them is selected as the
cluster head. When an event occurs, the cluster head
aggregates readings and MACs from its cluster nodes
(include itself) and generates the final reports, and then
forward these reports to the sink through forwarding nodes.
The sink has sufficient computing power and memory
space. Fig. 1 illustrates the system model of the sensor
network.

2436
Figure 1. the system model of the sensor network: the
circles outline the regions of clusters. CHs are the cluster
header and Rs are the forwarding nodes.
2.2 Threat model
Due to cost constraints sensor nodes are not
tamper-resistant and the attackers can compromise a node
through the radio communication channel or even
physically capture a node. We assume the sink employs
advanced defensive measures and cannot be compromised.
Adversaries have full control of compromised nodes and
can utilize these nodes to launch attacks. Same as other
schemes, we assume the sensor network has a short safe
phase after deployment. During this phase, no node is
compromised and it is safe to initialize sensor nodes and
distribute authentication keys.
3 The NLFS scheme
NLFS includes four phases: initialization phase, report
generation phase, en-route filtering phase and sink
verification phase.
3.1 Initialization phase
Before deployment, every node is preloaded with necessary
materials to establish symmetric keys. For example, if the
key management scheme RSDTMK [15] is employed,
nodes will be loaded with a master key and some functions.
After deployment, each node gets its ID and establishes
symmetric keys with its neighbors. The symmetric key is
used to encrypt reports transmitted between two nodes. In
this paper, A transmits a report R to B, which means R has
been encrypted.
The sink maintains a global pool of authentication keys
G= {KAi: 0 ≤ i ≤ N-1} and a cluster distribution table.
Table.1 shows the content of each row in the cluster
distribution table. Each row represents a cluster. The IDs
and keys are corresponding.
Table 1. The cluster distribution table
Cluster ID Node ID Key
CID1 ID1 ,ID2, ID3,ID4 KA1 , KA2 ,KA3 , KA4
CID2 ID5,ID6,ID7 KA5 , KA6 , KA7
… … …
CIDi IDi , IDj KAi , KAj
3.1.1 Initialization of node ID and key
After deployment, sensor nodes are organized into clusters
and each cluster generates a unique cluster ID. The Sink
creates a temporary variable SynID and initializes it to 0.
Then each cluster head follows steps below to obtain IDs
and keys of its cluster nodes. Initially, the cluster
distribution table is empty.
Step1: The cluster head sends report RUP {CID, L} to
the sink. CID and L denote the cluster id and the size of the
cluster respectively.
Step2: When the sink receives the report RUP, it
performs the following operations:
1. Select L authentication keys (KA1, KA2, …, KAl)
that have not been selected by other nodes from G;
2. Synchronize SynID to SynID +L;
3. Add a record {CID; SynID, SynID-1, …,
SynID-L+1; KA1, KA2, …, KAl} to the cluster
distribution table;
4. Send report RDOWN {CID; SynID; KA1, KA2, …,
KAl} to the cluster.
Step3: When the cluster head receives the report RDOWN,
it generates L pairs of key and ID: <SynID, KA1>,
<SynID-1, KA2>, …, <SynID-L+1, KAl>, and
distributes them to its cluster nodes.
Fig.2 illustrates the interactions between the sink and a
cluster. After the initialization, every node gets its ID and
key.
Figure 2. the interactions between the sink and a cluster
containing three nodes
3.1.2 Authentication key distribution
To make NLFS adaptive to highly dynamic networks, we
adopt the solution proposed in [11]. Every forwarding node
forwards the key distribution report RK to its q most
possible next hop nodes.
The detailed procedures for key distribution process are
as follow:
Step1: The cluster head collects IDs and authentication
keys from its cluster nodes and generates report RK {CID;
L; LS; ID1, KA1; ID2, KA2; …; IDi, KAi}, where LS is the
number of pairs of ID and key contained in RK and L is the
size of the cluster.
Step2: The cluster head forwards RK to its q forwarding
nodes.

2437
Step3: When a forwarding cluster header receives RK, it
performs the following operations:
1. Check if its cluster nodes have stored IDs and
keys in RK. If does, it deletes these IDs and keys from
RK and updates LS.
2. Create node set F. The elements of F are nodes in
the forwarding cluster that have not stored any IDs and
keys of nodes in the source cluster. The size of set F is
LF.
3. Compare LF and LS. If LF ≤ LS, select LF pairs of
ID and key from RK and distribute them to nodes in F. If
LF > LS, select LS nodes from F and distribute IDs and
keys in RK to them.
4. Delete IDs and keys that have been distributed
from RK and update LS. If LS > 0, turn to step2.
Each node in the forwarding cluster stores the ID and
key distributed by its cluster header, and also stores CID
and L of the source cluster. Fig.3 uses an example to show
the process of key distribution where q=1.The key
distribution method guarantees that no two keys stored in
one node are from the same cluster.
Figure 3. the distribution of keys of a cluster: the source
cluster contains four nodes (V1, V2, V3, CH1), and the
report RK sent to CH2 contains IDs and keys of nodes in the
source cluster. CH2, V5 and V4 select a pair of ID and key
in RK respectively and the rest ID and key (ID2, KA2) are
sent to CH3.
3.2 Report generation
When an event occurs, the cluster header collects sensor
data from sensing nodes in its cluster. Then it generates an
accurate description E of the event and forwards E back to
the sensing nodes. If the sensing node agrees on E, it
computes a MAC over E using its authentication key and
then sends the MAC and its ID to the Cluster header. The
report R the Cluster header finally generated is as follow:
R :{ E; CID; ID1, ID2, …, IDt; M1, M2, …, Mt}
During the en-route filtering phase, if the correctness of
a MAC in R is confirmed by a forwarding node F, the MAC
and its corresponding ID will be removed from R, and a
new MAC computed by node F and ID, CID of node F will
be added to R. The report R generated by a forwarding node
F is as follow:
R: {E; CID1; ID1,…, IDn; M1,…, Mn; CIDF; IDF; MF}
To provide high security, NLFS requires that at least r
(0< r ≤ t) nodes in R are from the same cluster.
3.3 En-route filtering
NLFS initializes serial ID numbers for nodes in the same
cluster. Therefore if the size of a cluster C is L, any two
nodes M, N in C should satisfy the following inequality:
|IDM - IDN| < L
When a forwarding node F receives report R, it mainly
performs the following verifications:
Node Legitimacy verification: For any node S that
endorses the report R, if F has the information of a node P
whose cluster ID is the same as S’, it computes |IDP – IDS|.
If |IDP – IDS| < L, we consider node S is legitimate, where L
is size of the cluster which p belongs to.
MAC verification: If node F stores a key that belongs
to a node in report R, it computes a new MAC over E using
this key. If the new MAC is the same as the corresponding
MAC in R, we consider the MAC in R is valid.
When a forwarding Cluster header (CH) receives report
R, it distributes R to its cluster nodes. Each cluster node F
then performs the following operations:
Step1: Check if R is in correct format and contains at
least t different IDs. Inform the CH to drop R otherwise.
Step2: Check if R contains at least r different IDs that
are in the same cluster. Inform the CH to drop R otherwise.
Step3: If node F has any of the CIDs in R, it checks the
legitimacy of corresponding nodes. If all these nodes are
legitimate, it sends their IDs to the CH as a report RL
{ID1, …, IDi}. Inform the CH to drop R otherwise.
Step4: If node F has any of the IDs in R, it checks the
correctness of corresponding MACs. If a MAC MS is
confirmed correct, F computes new MAC MF using its own
authentication key and sends report RM{IDS, MS; IDF, MF}
to the CH. Inform the CH to drop R, if there are invalid
MACs. If no ID in R is stored in F, F also computes MAC
MF and sends report RMF {IDF, MF} to the CH.
The CH collects verification reports from its cluster
nodes and decides whether to forward the report or not. The
detailed procedures are described as follow:
Step1: If the CH receives the message of dropping R, it
drops R.
Step2: If the number of reports RL is less than r or they
do not cover all nodes in R, the CH drops R. Step 2 checks
whether the legitimacy of every node endorses the report R
is verified.
Step3: If the CH does not receive any reports RM, it
forwards R to next hop. Otherwise, for each report RM, it
removes IDS, MS from R and adds IDF, MF to R.
Step4: The CH checks if R meets the requirements of t
and r. If not, it adds the IDs and MACs in reports RMF to R
and forwards R to next hop.
3.4 Sink verification
Since the sink has the cluster distribution table, it can
verify all MACs and IDs in R.

2438
4 Security Analysis
4.1 Filtering capacity
Theorem1. In NLFS, a false report injected by t-1
compromised nodes can be dropped within a / b hops.
Where a, b are the size of the largest cluster and the smallest
cluster.
Proof: According to the key distribution method, only
when a forwarding cluster cannot store all keys of the
source cluster, it forwards the rest keys to next hop.
Therefore, keys of the source cluster will be stored in
clusters that are with a/ b hops from the source cluster. Thus,
if totally t-1nodes are compromised, the attacker has to
forge a MAC and the corresponding ID. Since NLFS
requires a forwarding cluster verifies the legitimacy of all
IDs in R, report with invalid ID can be dropped
immediately. Even if the attacker can provide valid ID, the
report with forged MAC can also be dropped within a / b
hops; since the key corresponding to the valid ID must be
stored in a node within a/b hops.
4.1 Compromise tolerance
In NLFS, forwarding nodes not only verify the MACs but
also the legitimacy of nodes. In order to inject false reports
that NLFS cannot detect, the attacker has to compromise t
nodes in the same cluster. As illustrated in Fig.3 where t=3
and r =2, nodes C1, C2, C3, C4, C5 have been compromised,
a forged report generated by any three nodes of them can be
filtered since they are in different clusters and the forged
report cannot meet the requirement of r. Assume node S1
has also been compromised, a forged report R {e; CID1; C1,
S1; M1, MS; CID2; C2; M2} is injected through C1, where
nodes C1, S1 are in the same cluster and the report R is in
correct format . But, it can also be filtered by forwarding
nodes of node C1, since node C2is not legitimate.
Figure 4. collaborative false data injection attack launched
by compromised nodes
5 Simulation results
We study the performance of NLFS by simulation and
compare it with SEF [4] and NFFS [13] in terms of filtering
capacity, compromise tolerance. We simulate a 200*200
m2
field, where 1000 nodes are randomly deployed. The
transmission range of each node is 20m. The values of
security parameters t and r are 5 and 3 respectively.
Fig.4 illustrates how the percentage of false reports
filtered increases as the number of traveled hops grows.
From Fig.4 we can see the filtering probability of each
scheme increases, when the number of traveled hops
increases. Furthermore, NLFS has a higher false data
filtering capacity. It can drop 95% of false reports within 3
hops, while NFFS needs 10 hops and SEF needs more than
20 hops to achieve the same filtering result.
Figure 5. the percentage of false reports filtered as a
function of the number of hops they traveled
Fig. 5 illustrates how many compromised nodes every
scheme can tolerate. Fig.5 does not show the results of SEF,
since it can tolerate less than 15 nodes. The simulation
results are averaged over 1000 random tests.
Figure 6. the probability to break NLFS as a function of the
number of compromised nodes
Fig.6 illustrates how the percentage of false reports
filtered increases as the number of compromised nodes
grows. In NFFS, attackers have a great probability of
injecting false reports that can only be filtered by the sink
through nodes closer to the sink, since these nodes hold
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
The number of traveled
Percentageoffalsereportsfiltered
SEF
NFFS
NLFS
0 50 100 150 200 250 300
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
The number of compromised nodes
theprobabilitytobreakthesecuritydesign
NFFS
NLFS

2439
fewer auth-keys than nodes closer to clusters. Therefore, as
the number of compromised nodes increases, the filtering
capability of NFFS decreases quickly, while NLFS
decreases much more slowly.
Figure 7. The percentage of false reports filtered as a
function of the number of compromised nodes.
6 Conclusion
In this paper, we propose a node legitimacy based scheme
to address false data injection attack. Our novel solution is
to initialize serial ID numbers for nodes in the same cluster
and verify the legitimacy of nodes using IDs stored in
forwarding nodes. Compared with others, NLFS can filter
false reports much earlier and can tolerate more
compromised nodes, since it verifies not only MACs in the
reports, but also the legitimacy of nodes. However, NLFS is
more complicated than SEF and NFFS in the en-route
filtering phase and we will further improve the scheme and
make it more resilient and efficient.
REFERENCES
[1] H.D. Wang, Q. Li, Achieving robust message
authentication in sensor networks: a public-key based
approach. WIRELESS NETWORKS, 16(4), 2010,
999-1009.
[2] Y. Zhang, W. Liu, W. Lou, Location-based
compromise-tolerant security mechanisms for
wireless sensor networks. IEEE Journal on Selected
Areas in Communications, 24(2), 2006, 247-260.
[3] R. Lu, X. Lin, H. Zhu, BECAN: a bandwidth-efficient
cooperative authentication scheme for filtering
injected false data in wireless sensor networks. IEEE
Transactions on Parallel and Distributed Systems,
23(1), 2012, 32-43.
[4] F. Ye, H. Luo, S. Lu, L. Zhang, Statistical en-route
filtering of injected false data in sensor networks.
IEEE Journal on Selected Areas in Communication,
23(4), 2005, 839-850.
[5] S. Zhu, S. Setia, S. Jajodia, An interleaved
hop-by-hop authentication scheme for filtering of
injected false data in sensor networks, ACM
Transactions on Sensor Networks(TOSN), 3(3), 2007.
[6] M Ma, Resilience of sink filtering scheme in wireless
sensor networks, Computer Communications, 30(1),
2006, 55-65.
[7] E. AYDAY, F. DELGOSHA , F. FEKRI,
Location-aware security services for wireless sensor
networks using network coding, Proc. 26th IEEE
Conf. on Computing and Communicating , Alaska,
USA, 2007, 1226−1234.
[8] K. Ren, W.J. Lou, Y.C. Zhang, LEDS: Providing
location-aware end-to-end data security in wireless
sensor networks, IEEE Transactions on Mobile
Computing, 7(5), 2008, 585-598.
[9] F. Yang , X. Zhou, Q. Zhang, Multi-dimensional
resilient statistical en-route filtering in wireless sensor
networks, Proc. 5th International Conf. on Grid and
Pervasive Computing, Hualien, TAIWAN, 2010,
130-139.
[10] Nghiem, T.H. Cho, A multi-path interleaved
hop-by-hop en-route filtering scheme in wireless
sensor networks, Computer Communications, 33(10),
2010, 1202-1209.
[11] Z. Yu, Y. Guan, A dynamic en-route filtering scheme
for data reporting in wireless sensor networks,
IEEE/ACM Transactions on Networking, 18(1), 2010,
150-163.
[12] Z. LIU, J. WANG, Geographical information based
false report filtering scheme in wireless sensor
networks, Journal of Communications, 33(2), 2012,
156-163.
[13] J. Wang, Z. Liu, S. Zhang and X. Zhang, Defending
collaborative false data injection attacks in wireless
sensor networks, Information Sciences, 254, 2014,
39-53.
[14] P. Kuila, P.K. Jana, A novel differential evolution
based clustering algorithm for wireless sensor
networks, Applied soft computing, 25, 2014, 414-425.
[15] F. Gandino, B. Montrucchio, M. Rebaudengo, Key
management for static wireless sensor networks with
node adding, IEEE Transaction on industrial
informatics, 10(2), 2014, 1133-1143.
0 20 40 60 80 100 120 140 160 180
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
The number of compromised nodes
Thepercentageoffalsedatafiltered
NFFS
NLFS

Node Legitimacy Based False Data Filtering Scheme in Wireless Sensor Networks

More Related Content

What's hot (16)

Viewers also liked (20)

Similar to Node Legitimacy Based False Data Filtering Scheme in Wireless Sensor Networks (20)

More from Eswar Publications (20)

Recently uploaded (20)

Node Legitimacy Based False Data Filtering Scheme in Wireless Sensor Networks