On Secure Communication over Wireless Sensor Networks
Stefan Rass
1
and Michał Koza
2
1
Institute of Applied Informatics, Universit¨at Klagenfurt, Universit¨atsstrasse 65-67, 9020 Klagenfurt, Austria
2
Faculty of Fundamental Problems of Technology, Wrocław University of Technology, Wrocław, Poland
Keywords:
Wireless Sensor Network, Security: Secrecy, Secret Sharing, Perfectly Secure Message Transmission.
Abstract:
This paper investigates (perfectly) secure message transmission over a wireless sensor network. Using a
layered network architecture and a very simple form of routing, we show how to construct an arbitrarily secure
communication channel over a given infrastructure of wireless devices. Our construction is computationally
cheap and requires no cryptographic primitive beyond symmetric encryption on the channels. The security of
the transmission can be made arbitrarily strong (in an information-theoretic sense).
1 INTRODUCTION
Secret message transmission is doubtlessly an impor-
tant aspect in vast amount of communication scenar-
ios, and has been thoroughly studied in the wired do-
main for decades. The increasing interest and expand-
ing range of applications in the wireless domain sheds
a different light on security requirements when de-
signing wireless transmission protocols. In this work,
we propose a simple protocol for transmission over a
wireless infrastructure, which explicitly exploits fea-
tures of the wireless domain to create a highly secure
channel through an otherwise not trustworthy envi-
ronment. The limited capabilities of small wireless
(sensor) devices, such as computational and battery
power or the physical access protection of crypto-
graphic keys, can become and obstacle for using ad-
vanced cryptography. So, we are after slim and com-
putationally cheap protocols.
Fortunately, some features of wireless networks
can be used in order to construct a highly secure
transmission channel. Devices are usually simple and
small, so it can be very challenging for an adversary
to physically locate them in an environment. Such
devices are often also cheap and can be deployed in
large numbers, thus facilitating the use of secret shar-
ing and/or multi-path transmission.
Intuitively, if we can not make the transmission
secure, why not try to make it “invisible”? This it the
road taken in the recent paper of (Klonowski et al.,
2011), and our goal in this work is further exploring
that direction by devising an extended version of their
protocol, hereafter referred to as BasicSWTP. We re-
fer the reader to (Klonowski et al., 2011) for full de-
tails, and will give only a brief review in section 2.
Organization of the Paper. Preliminaries and
groundwork preparation done in Section 2. Section
3 is devoted to a presentation of BasicSWTP and its
extension, along with security proofs. Conclusions
follow in Section 4.
Related and Previous Work. A lot of papers dis-
cuss secret message transmission in wireless net-
works based on symmetric cryptography and key pre-
distribution (e.g. (Perrig et al., 2004; Jaworski et al.,
2009; Cicho´n et al., 2009; Anderson et al., 2004;
Chan et al., 2003; Du et al., 2005)). The main goal
of these protocols is establishing secure channels be-
tween devices, where channels are secured by sym-
metric cryptography. While key managementand pre-
distribution Although our scheme can (and should)
be used with key-predistribution schemes (Klonowski
et al., 2007; Ren et al., 2006), the problem stated in
this paper cannot be solved by symmetric schemes
with key-predistribution only. Nevertheless, the re-
sults in (Chan and Perrig, 2005; Eschenauer and
Gligor, 2002) are notable and related and are an as-
set to our work. Most closely related to ours is the
work of (Di Pietro et al., 2006). Our contribution tar-
gets the same goals, yet in addition comes with an
information-theoretic analysis, which is not provided
in previous related work.
364
Rass S. and Koza M..
On Secure Communication over Wireless Sensor Networks.
DOI: 10.5220/0004018403640367
In Proceedings of the International Conference on Security and Cryptography (SECRYPT-2012), pages 364-367
ISBN: 978-989-8565-24-2
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
2 PRELIMINARIES
Our network model is the same as in (Klonowski
et al., 2011), which is a wireless multi-hop sensor net-
work. In graph-theory language, the network model
graph G(V, E) is undirected and t-partite. The (pair-
wise disjoint) partition V = L
0
˙
L
1
˙
...
˙
L
t1
is such
that no two nodes within the same layer L
i
talk to
each other, and every node from L
i
has a logical link
(i.e. a symmetrically encrpted channel) to every node
in L
i+1
for i = 0, 1,.. . ,t 2. One way of building
such a layered architecture is using one of many ex-
isting clustering protocols (see (Yuand Chong, 2005))
or general key-distribution protocols (e.g. Diffie-
Hellman or (Perrig et al., 2004)). We assume low
level protocol services like cluster level routing, col-
lision avoidance etc., as provided by the network.
The Basic Transmission Protocol BasicSWTP. On
each layer L
i
the message m appears split into l parts,
which are created by the sender s L
0
using XOR-
secret-sharing. A relay node in L
i
(i 1) takes its l
incoming share parts, combines them to get its mes-
sage share s
i
, invokes another sharing on it and for-
wards the new parts to l nodes in L
i+1
. The crux of
(Klonowski et al., 2011)’s construction is a special in-
dexing scheme that pseudorandomly determines the
receivers in the next layer so that all relay nodes
within L
i
choose the same set of l receivers within
L
i+1
. This assures that the subgraph of G that is
used for the actual payload delivery always remains t-
partite subgraph, with each intermediate layer having
cardinality l. The parameter l controls the number of
shares, and hence the number of paths in the transmis-
sion, and is called the forking parameter (Klonowski
et al., 2011). Figure 1 shows an example.
Adversary Model. For security, assume that the ad-
versary corrupts no more than K nodes in total, where
a corrupted node has all its content (including any
cryptographic keys) exposed to the attacker. Further-
more, we assume the adversary’s computational pow-
ers sufficiently constrained to inhibit a ciphertext only
attack on the symmetric encryption used for any wire-
less transmission in the network. This assumption
is necessary to rule out trivialities by situations in
which the adversary gets all the information flowing
between the sender and the receiver. Notice that this
sort of computational intractability assumption might
indeed permit breaking a public-key encryption.
Security Model. The symbol Π denotes a gen-
eral message transmission protocol. With H(·) and
H(·|·), we denote the unconditional and conditional
Shannon-entropy.
Definition 2.1 ((Franklin and Wright, 2000)). Take
ε > 0, and let Π be a message transmission proto-
col. Suppose that for conveyance of a message M
{0,1}
, the packets C
1
,...,C
n
{0,1}
are transmit-
ted over the network (constituting the protocol’s tran-
script). The adversary’s view on the transmission of
M is adv(M) {C
1
,...,C
n
}. A protocol ε-secure, if
H(M|adv(M)) {0,H(M)}, and
P(H(M|adv(M)) = 0) ε,
i.e. the adversary can disclose M with a chance of at
most ε. We call the protocol Π efficient, if the size of
the transcript, i.e.
n
i=1
|C
i
|, is polynomial in the size
of the message M, the size of underlying network (in
terms of nodes), and log
1
ε
. A protocol that is ε-secure
for any ε > 0 is said to enjoy perfect secrecy.
Security of BasicSWTP. The following two results
are found in (Klonowski et al., 2011) and regard the
security of the above sketched transmission scheme:
Theorem 2.1. The strategy of putting all corrupted
nodes in exactly one layer maximizes the probability
of corrupting the message.
Corollary 2.2. In the system with forking parameter
l, with n nodes in each layer, and the adversary capa-
ble of corrupting K nodes, the probability p that the
adversary discloses a secret message is
p =
0, if K < l;
K
l
/
n
l
, if l K < n;
1, if K n.
(1)
The Concept of Vulnerability. For analyzing secu-
rity, we can set up the transmission as an attacker-
defender-game in the obvious way: the defender
(message sender) chooses the t-partite subgraph of G
for transmission (grayish nodes in figure 1), while the
attacker chooses his nodes to conquer. The game’s
objective is secret content delivery, so the sender
(player 1) scores 1 every securely delivered message,
and receives zero payoff in case of a successful mes-
sage disclosure by the attacker (player 2). The game
is zero-sum, meaning that the attacker scores 1 for ev-
ery disclosed message and zero otherwise. With this
setting, standard techniques of game-theory let us de-
termine the optimal behavior for both players, with
the optimal strategies of the attacker being character-
ized by theorem 2.1 already. The optimal revenue for
player 1, i.e. the average success-rate of secret mes-
sages delivered under the hypothesis that the attacker
acts optimal too is the game’s saddle-point value v.
See (Rass and Schartner, 2010) for full details on this
game-theoretic approach. The important fact here is
that corollary 2.2 and the details in (Klonowski et al.,
OnSecureCommunicationoverWirelessSensorNetworks
365
2011) already give us the optimal transmission and at-
tack strategies, along with equation (1) providing the
game’s saddle-point value.
3 SECURE TRANSMISSION
It is evident that choosing a larger forking parameter
occupies more of the network and increases the band-
width demand. Hence, lower values of l are desirable,
yet are paid for with less security.
Framework Protocol. Our proposed extension is
running BasicSWTP with a small (fixed) forking pa-
rameter on a set of shares to the message, rather than
on the message as such. Initially, the message m is
shared as m = r
1
r
2
··· r
k
(where is the bit-
wise XOR). Each r
i
for i = 1,2,...,k is transmitted
using BasicSWTP with an either chosen or prescribed
(small) forking parameter l. The reconstruction by the
receiver happens in the obvious way. This frame-
work protocol”, sketched in figure 1 as the wrapper
around BasicSWTP, is indeed secure and efficient:
Theorem 3.1. Let the channel be characterized by
the number n of nodes per layer and let the protocol
work with a fixed forking parameter l n. Assume
that a sender estimates the threshold of the adversary
as K n. Then, in the sense of Definition 2.1, arbi-
trarily secure communication is efficiently possible if
and only if K < n.
Proof sketch: Corollary 2.2 gives the likelihood
p of secret delivery in a single round, and the XOR-
sharing enforces the attacker to catch all shares when-
ever the message shall be disclosed. So, if p = 1, then
the chance to catch all shares is 1 and the transmission
is insecure. On the contrary, if p < 1, then the chance
to catch all of the shares decays exponentially fast, so
we can choose k large enough to lower this chance be-
low any chosen ε > 0. The bit-complexity is k· l · |m|,
where |m| is the length of the message. The overall
overhead is O
|m| · l · log
1
ε
, i.e. polynomial.
Secure Key-exchange. We can dress up our claims
in information-theoretic terms too, if we wrap the
framework protocol around BasicSWTP, in order to
exchange keys with assured entropy. The following is
a technical gadget, needed to establish the second of
our results, which is theorem 3.3.
Theorem 3.2 ((Rass and Schartner, 2010)). Let the
sender emit secret messages M of entropy H(M), and
let C denote the information that the adversary can
acquire by eavesdropping. Then the cross-entropy
I(M;C), i.e. the amount of information that leaks
out the channel satisfies I(M;C) ρ · H(M), where
message
sharingsharing
r
1
,. .. ,r
k
layer L
i1
L
i
L
i+1
BasicSWTP with forking parameter
l = 3 to convey r
1
,. .. ,r
k
Figure 1: Example illustration of our protocol.
ρ = maxP(successful attack), and the maximum is
taken over all randomized actions of the attacker.
It should be noted that the value ρ in theorem 3.2
can be obtained using game-theoretic tools (Rass and
Schartner, 2010) or directly from Corollary 2.2.
Theorem 3.3. Let a source with entropy h generate a
sequence of independent random numbers r
1
,...,r
k
,
which are transmitted via BasicSWTP. Let the sender
and receiver derive a common secret κ = r
1
r
2
··· r
k
. Then, even for a computationally unbounded
adversary, the secret has entropy H(κ) (1 ρ
k
)h,
where ρ is given by Corollary 2.2 applied to the in-
stantiation of BasicSWTP.
Proof sketch: Let the sender transmit k indepen-
dent uniform random numbers r
1
,...,r
k
and write
κ = r
1
r
2
··· r
k
for the exchanged key κ. By
the stochastic independence, the attack probability
is by Corollary 2.2 no larger than p
k
. Applying
Theorem 3.2 with p
k
bounds the leaking informa-
tion I(C;κ) p
k
· H(κ), where C denotes the infor-
mation that the adversary gains from eavesdropping.
By definition, I(C;κ) = H(κ) H(κ|C), so we get
H(κ|C) (1 p
k
) · H(κ).
4 CONCLUSIONS
While (Klonowski et al., 2011)s protocol, here called
BasicSWTP, achieves good security based on a free
choice of the forking parameter, having multiple such
channels run concurrently can quickly occupy and
congest the infrastructure. For that matter, it appears
reasonable to use another sharing on top of Basic-
SWTP, which lets us work with smaller forking pa-
rameters while not too badly affecting the security (as
theorems 3.1 and 3.3 demonstrate).
Our protocol is simple and creates an arbitrary
secure channel through a potentially hostile environ-
ment. Moreover, it is provably secure and efficient,
SECRYPT2012-InternationalConferenceonSecurityandCryptography
366
and does not rest on computational intractability as-
sumptions beyond what is needed to establish se-
cure symmetrically encrypted channels. Even from
an information-theoretic point of view, it is possible
to use our scheme for secret key-exchange only, so
as to gain security even against a computationally un-
bounded adversary. Moreover, our scheme is light-
weight in the sense of imposing little computational
effort within each relay node. So it can be imple-
mented in cheap and power-limited devices, partic-
ularly small sensor-devices. The protocol offers two
degrees of freedom which lets us control the commu-
nication overhead and find a suitable balance between
security and communication overhead. This facili-
tates a fair-use policy of the channel, if multiple ses-
sions run concurrently over the same set of devices.
In future work, we will report on empirical evaluation
via simulation.
Unfortunately, our protocol is vulnerable to
denial-of-service attacks in its present form. The
XOR-sharing as we used is most vulnerable to cor-
rupted shares, since the secret is unrecoverable if one
share is lost or modified. However, XOR-sharing is
not a must and might be replaced by conventional
polynomial sharing that comes with better error cor-
recting facilities (e.g. Shamir’s scheme). Guarding
against loss of shares and routing attacks are subject
of future research.
This work is written with wireless networks in
mind, but the presented algorithm can work in any
network where the construction of the aforemen-
tioned layered architecture is possible. In particular,
it seems possible to apply it in wired networks like
cascades of mix servers presented in (Klonowski and
Kutylowski, 2005).
ACKNOWLEDGEMENTS
The paper was partially supported by grant S10026
from the Institute of Mathematics and Computer Sci-
ence of the Wroclaw University of Technology.
REFERENCES
Anderson, R., Chan, H., and Perrig, A. (2004). Key infec-
tion: Smart trust for smart dust. In Proc. of IEEE Int.
Conf. on Network Protocols (ICNP).
Chan, H. and Perrig, A. (2005). PIKE: peer intermediaries
for key establishment in sensor networks. In INFO-
COM 2005. Proc. of the 24th Annual Joint Conf. of
the IEEE Computer and Communications Societies.,
volume 1, pages 524–535.
Chan, H., Perrig, A., and Song, D. (2003). Random key pre-
distribution schemes for sensor networks. In Proc. of
the IEEE Symp. on Security and Privacy, pages 197–
213. IEEE Computer Society.
Cicho´n, J., Grza´slewicz, J., and Kutyłowski, M. (2009). Key
levels and securing key predistribution against node
captures. In ALGOSENSORS, pages 64–75.
Di Pietro, R., Mancini, L. V., and Mei, A. (2006). Energy
efficient node-to-node authentication and communica-
tion confidentiality in wireless sensor networks. Wirel.
Netw., 12:709–721.
Du, W., Deng, J., Han, Y. S., Varshney, P. K., Katz, J.,
and Khalili, A. (2005). A pairwise key predistribution
scheme for wireless sensor networks. ACM Trans. Inf.
Syst. Secur., 8(2):228–258.
Eschenauer, L. and Gligor, V. D. (2002). A key-
management scheme for distributed sensor networks.
In CCS ’02: Proc. of the 9th ACM Conf. on Computer
and communications security, pages 41–47. ACM
Press.
Franklin, M. and Wright, R. (2000). Secure communica-
tion in minimal connectivity models. J. of Cryptology,
13(1):9–30.
Jaworski, J., Ren, M., and Rybarczyk, K. (2009). Random
key predistribution for wireless sensor networks using
deployment knowledge. Computing, 85(1–2):57–76.
Klonowski, M., Koza, M., and Kutyłowski, M. (2011). How
to transmit messages via WSN in a hostile environ-
ment. In Proc. of the Int. Conf. on Security and Cryp-
tography (SECRYPT), pages 134–143. SciTePress.
Klonowski, M. and Kutylowski, M. (2005). Provable
anonymity for networks of mixes. In Information Hid-
ing, pages 26–38.
Klonowski, M., Kutyłowski, M., Ren, M., and Rybarczyk,
K. (2007). Forward-secure key evolution in wireless
sensor networks. In CANS, pages 102–120. Springer.
Perrig, A., Stankovic, J. A., and Wagner, D. (2004). Se-
curity in wireless sensor networks. Commun. ACM,
47(6):53–57.
Rass, S. and Schartner, P. (2010). A unified framework
for the analysis of availability, reliability and security,
with applications to quantum networks. IEEE Trans.
on Systems, Man, and Cybernetics – Part C: Applica-
tions and Reviews, 40(5):107–119.
Ren, M., Das, T. K., and Zhou, J. (2006). Diverging keys in
wireless sensor networks. In ISC, pages 257–269.
Yu, J. Y. and Chong, P. H. J. (2005). A survey of clustering
schemes for mobile ad hoc networks. Communica-
tions, 7(1):32–48.
OnSecureCommunicationoverWirelessSensorNetworks
367