QUANTIFYING MISBEHAVIOUR ATTACKS AGAINST THE
SELF-ORGANIZED PUBLIC KEY MANAGEMENT ON MANETS
Eduardo da Silva, Aldri Luiz dos Santos, Luiz Carlos Pessoa Albini
NR2/LARSIS − Department of Informatics, Federal University of Paran´a, Curitiba, Brazil
Michele N. Lima
Laboratoire d’informatique de Paris 6, Universit´e Pierre et Marie Curie, Paris, France
Keywords:
MANET, Self-Organized Public Key Management, Quantification.
Abstract:
Among the key management schemes for MANETs, the Self-Organized Public Key Management System
(PGP-Like) is the main chaining-based key management scheme. It is fully self-organized and does not
require any certificate authority. Two kinds of misbehavior attacks are considered to be great threats to PGP-
Like: the impersonating and the lack of cooperation attacks. This work quantifies the impact of such attacks
on PGP-Like. Simulation results show that PGP-Like was able to maintain its effectiveness when submitted
to the lack of cooperation attack, contradicting previously theoretical results. It correctly works even in the
presence of more than 60% of misbehaving nodes, although the convergence time was affected with only 20%
of misbehaving nodes. On the other hand, PGP-Like was completely vulnerable to the impersonating attack.
Its functionality is affected with just 5% of misbehaving nodes, confirming previously theoretical results.
1 INTRODUCTION
Due to the lack of infrastructureand dynamic environ-
ment, Mobile Ad Hoc Networks (MANETs) are ex-
tremely vulnerable to active and passive attacks (Dje-
nouri et al., 2005). Units of such networks are mo-
bile and independent from each other, making net-
work management and security critical tasks. Further-
more, units can have malicious or selfish behavior, or
even be compromised by adversaries. Indeed, tradi-
tional security protocols do not correctly fit into the
paradigm of MANETs.
Cryptographyis the main technique used to ensure
data communication security. It provides information
integrity, authenticity, non-repudiation and confiden-
tiality. Cryptographic algorithms require the use of
pair-wised keys. The secure administration of these
keys, known as key management, must consider gen-
eration, storage, distribution, protection and revoca-
tion of the keys, and also ensures availability to au-
thentic units (nodes).
Traditional cryptographic systems have been di-
vided into symmetric and asymmetric ones, depend-
ing on the way they use keys. Although symmet-
ric systems require less processing than asymmetric
ones, they are not scalable, demanding that secret
keys must be shared either by a secure pre-established
channel or before network formation. Therefore,
symmetric schemes are difficult to be applied on
MANETs (Chlamtac et al., 2003). On the other hand,
traditional asymmetric systems require a trusted en-
tity to authenticate certificates and keys. However,
establishing a trusted entity in a MANET is a chal-
lenge, due to their decentralized organizationand lack
of trust model (Butty´an and Hubaux, 2003).
Key management for MANETs must deal with
dynamic topology and be self-organized and decen-
tralized (Hegland et al., 2006;
ˇ
Capkun et al., 2006;
van der Merwe et al., 2007). It must also satisfy re-
quirements like: (i) not having a single point of fail-
ure; (ii) being compromise-tolerant, meaning that the
compromise of a certain number of nodes does not af-
fect the security between the non-compromised ones;
(iii) being able to efficiently and securely revoked
keys of compromised nodes, and update keys of non-
compromised nodes; (iv) being efficient in terms of
storage, computation, and communication.
Several key management schemes for MANETs
can be found in the literature. Among them, the Self-
Organized Public Key Management System (Hubaux
128
da Silva E., Luiz dos Santos A., Carlos Pessoa Albini L. and N. Lima M. (2008).
QUANTIFYING MISBEHAVIOUR ATTACKS AGAINST THE SELF-ORGANIZED PUBLIC KEY MANAGEMENT ON MANETS.
In Proceedings of the International Conference on Security and Cryptography, pages 128-135
DOI: 10.5220/0001922701280135
Copyright
c
 SciTePress

et al., 2001;
ˇ
Capkun et al., 2003) is the main chaining-
based key management scheme. From now on the
Self-Organized Public Key Management System will
be called PGP-Like. It is a self-organized public key
management scheme based on the PGP concepts, in
which all pair-wised keys are created by nodes them-
selves. Nodes also issue certificates to other ones
in which they trust. Each node has a local certifi-
cate repository that is periodically exchanged with its
neighbors, forming certificate chains. Two kinds of
misbehavior attacks are considered to be great threats
to PGP-Like, the impersonating and the lack of coop-
eration attacks (Engel et al., 2006). However, studies
of such attacks over PGP-Like found in the literature
are only theoretical (Yi and Kravets, 2004; He et al.,
2007; Gouda and Jung, 2004). There is no work in
the literature that quantifies the impact of these mis-
behavior attacks over PGP-Like.
This work quantifies the PGP-Like effectiveness
under these two different misbehavior attacks, the im-
personation and the lack of cooperation. The im-
personation attack, called Sybil, consists in creating
false identities able to be authenticated by PGP-like.
The lack of cooperation attack, called Blackhole, con-
sists in selfish nodes not cooperating with the net-
work. The worst case scenario for the PGP-Like key
management scheme is a small variance of the typical
Blackhole. In this variance, selfish nodes only misbe-
havior during certificate exchanges, working correctly
during all other network operations.
Simulation results show that PGP-Like is com-
pletely vulnerable to Sybil attacks. Its functional-
ity can be compromised even in the presence of very
few attackers, confirming the theoretical assessments
found in the literature. However, PGP-Like maintains
its effectiveness against blackhole attacks, almost in-
dependently from the number of attackers. This result
is completely different from the theoretical assess-
ments found in the literature (van der Merwe et al.,
2007; Wu et al., 2007).
The rest of this paper is organized as follows:
Section 2 briefly discusses the kinds of attacks on
MANETs; Section 3 describes the PGP-Like charac-
teristics, functionality and vulnerabilities; Section 4
contains the metrics used on the PGP-Like evalua-
tion; Section 5 presents the PGP-Like evaluation un-
der the Sybil and the Blackhole attacks; finally, Sec-
tion 6 draws the conclusions and future work.
2 ATTACKS OVER MANETS
MANETs are susceptible to many security issues re-
lated to their natural characteristics and properties.
Multihop communication, lack of infrastructure, lim-
ited resources and mobility make them vulnerable
to passive and active attacks, in which misbehaving
nodes can eavesdrop or delete packets, modify packet
contents or impersonate other nodes (Djenouri et al.,
2005).
Attacks in MANETs can be divided into modifi-
cation, fabrication, impersonating and lack of coop-
eration attacks (Michiardi and Molva, 2003). Among
these, the impersonation and the lack of cooperation
attacks are the ones that can cause more damage to
the PGP-Like key management scheme (Engel et al.,
2006). Thus, without loss on generality, this work fo-
cus only on these two kinds of attacks. Impersonating
attacks consist in using false identities to deceive net-
work protocols. In the lack of cooperation attacks,
selfish nodes use network resources but do not coop-
erate with any network operations.
Examples of impersonating attacks and lack of
cooperation attacks for MANETs are the Sybil
(Douceur, 2001) and the blackhole (Al-Shurman
et al., 2004) attacks, respectively. A large amount of
work can be found in the literature to deal with these
threats, for example: techniques to detect Sybil at-
tacks can be found in (Zhang et al., 2005; Douceur,
2001); techniques to detect and deal with blackhole
attacks can be found in (Agrawal et al., 2008; Ra-
maswamy et al., 2003). Even though these attacks
are considered dangerous threats for MANETs, the
analysis of their effects are mainly focused on rout-
ing protocols. All work found in the literature which
mention these attacks over PGP-Like are theoretical
(van der Merwe et al., 2007; Wu et al., 2007), or sim-
ply expected behavior. No one really quantifies the
behavior of PGP-Like under these attacks.
3 PGP-LIKE
PGP-Like is a public key management scheme that
uses certificate chains (
ˇ
Capkun et al., 2003; Hubaux
et al., 2001). Private and public keys of nodes are
created by the nodes themselves like PGP concepts
(Zimmermann, 1995). In addition, each node issues
public key certificates to other nodes it trusts. The
nodes themselves store and distribute certificates in a
self-organized manner.
On PGP-Like, public keys and certificates are rep-
resented by a directed graph G(V, E) (
ˇ
Capkun et al.,
2003), in which V represents the public keys of the
nodes and E represents the certificates. A directed
edge between two vertexes, K
u
and K
v
, represented by
(K
u
→ K
v
), denotes a signed certificate with the public
key of node u, which binds K
v
to node v. In addition,
QUANTIFYING MISBEHAVIOUR ATTACKS AGAINST THE SELF-ORGANIZED PUBLIC KEY MANAGEMENT
ON MANETS
129

a path connecting two vertex, K
u
and K
w
(K
u
K
w
),
represents a certificate chain from K
u
to K
w
. Note that,
in (K
u
K
w
), the first certificate on the chain can di-
rectly verified by node u, each remaining certificate
can be verified using the public key of the previous
certificate in the chain and the last certificate contains
the public key of node w.
In PGP-Like, if a node u believes that a given
public key K
v
belongs to a given node v, it can is-
sue a signed certificate binding K
v
to node v, denoted
by (v, K
v
)
prK
u
. Certificates are issued with a limited
validity time T
V
. Initially, node v keeps in its local
repositories only the certificates v issued and the cer-
tificates that other nodes issued to v, i.e., each time
node u issues a certificate that binds K
v
to node v,
u sends the certificate to v. Thus, each certificate is
stored at least twice, by u and v.
To correctly authenticate a node via a certificate
chain, a node must guarantee that all certificates on
the chain are valid and correct. To build appropri-
ate certificate chains each node u maintains two cer-
tificate repositories, the updated and the non-updated
repository (
ˇ
Capkun et al., 2003). The updated certifi-
cate repository, represented by G
u
, contains the sub-
set of certificates that node u maintains up-to-date,
i.e., node u requests updates for these certificates from
their issuers before they expire. The non-updated cer-
tificate repository, represented by G
N
u
, contains the
certificates collected by node u that have not been up-
dated yet, updated and also the expired certificates.
Figure 1 shows the update local certificate reposito-
ries of nodes u and v (Figure 1a and 1b, respectively).
Ks
Km
Kr
Kf
Kj
Ke
Kw
Kg
Ku
Kb
Kd
Kv
Ke
Kh
Kw
Kc
Kt
(b)
(a)
G
u
G
v
Figure 1: Certificate graphs.
Nodes also implement a certificate exchange mecha-
nism. It consists in periodically exchanging certifi-
cates with its physical neighbors, i.e., node u periodi-
cally multicasts its subgraphs, G
u
and G
N
u
, to its phys-
ical neighbors. Therefore, after several certificate ex-
changes and considering nodes mobility, all certifi-
cates might be stored by all nodes. The expected time
needed by a certificate to reach all nodes is called con-
vergence time (T
CE
).
When node u wants to verify the authenticity of
the public key K
v
of node v, they firstly merge their
updated certificate repositories, creating G
1
= G
u
∪
G
v
(Figure 2). Then, node u tries to find (K
u
K
v
) ∈
G
1
. If ∃(K
u
K
v
) ∈ G
1
, node u uses the certificates
on this path to authenticate K
v
. If ¬∃(K
u
K
v
) ∈ G
1
,
then u creates G
2
= G
u
∪ G
N
u
and tries to find (K
u
K
v
) ∈ G
2
. If such a path can be found, node u must
update all expired certificates, check their correctness
and authenticate K
v
. If ¬∃(K
u
K
v
) ∈ G
2
, node u
fails to authenticate K
v
.
Kw
Kc
Kh
G U G
u v
Ku
Ks
Km
Kr
Kf
Kj
Kg
Ke
Kt
Kv
Kb
Kd
Figure 2: Path in the merged certificate repositories.
Each certificate is issued with a validity time T
V
and it
can be revoked either by an explicit or by an implicit
revocation scheme. In the explicit revocation scheme,
the issuer node creates an explicit revocation state-
ment and sends it to each node that regularly updates
this certificate. The revocation statement might be re-
propagated until it reaches all nodes. Thus, the re-
vocation time might be up to T
CE
. On the other hand,
the implicit revocation scheme is based on the validity
time of the certificates. After a certificate expires, it is
stored in the non-update local certificate repositories
of the nodes, and considered invalid.
PGP-Like assumes the existence of a trust model
between nodes, and creates certificate chains, based
on it. However, creating these chains can take a long
time, as nodes must interact with each other to form
them. Thus, a node might not be able to authenticate
a certificate during system initialization. This char-
acteristic can be explored by several kinds of attacks,
like blackhole, in which selfish nodes can block the
propagation of issued certificates. Furthermore, cer-
tificate chains represent the trustworthiness between
nodes and are called trust chains. Note that trust
chains are weak authentications, as they assume that
trust is transitive, i.e., if node A trusts in node B, and
node B trusts in node C, then node A also trusts in
node C. An attacker can exploit this fact and (i) is-
sue a certificate trying to bind K
v
to node f; (ii) issue
a certificate trying to bind a false key K
′
v
to an au-
thentic node v; (iii) create a false identity m, create
a false key K
m
, issue a certificate that binds K
m
to m
and try to convince a valid user that this certificate is
valid. According to (
ˇ
Capkun et al., 2003), PGP-Like
prevents all these attacks by allowing nodes to detect
SECRYPT 2008 - International Conference on Security and Cryptography
130

inconsistent certificates and to determine which user-
key bindings are correct.
If node u receives a certificate which contains the
user-key binding (v, K
v
) and it does not contain this
binding in any certificate in G
u
or G
N
u
, then u classi-
fies this certificate as un-specified. Labeling a certifi-
cate as un-specified means that node u does not have
sufficient information to verify the authenticity of this
certificate. If node u receives another certificate with
the user-key binding (v, K
′
v
), it labels both certificates
as conflicting. If node u does not receive any con-
flicting certificate for a certain period of time it clas-
sifies the original certificate as non-conflicting. When
node u detects a conflict, it tries to find chains of non-
conflicting and valid certificates to the public keys K
v
and K
′
v
. Based on these chains node u might decide to
classify a certificate as non-conflicting and the other
one as false. If node u cannot reach any decision, both
certificates remain classified as conflicting.
This method can be easily implemented and it
might guarantee that an attacker cannot issue a cer-
tificate to bind K
v
to node f or vice versa. However,
it will not work correctly if the attacker creates sev-
eral false identities to itself and maintains a correct
behavior for a while (Yi and Kravets, 2004; He et al.,
2007; Gouda and Jung, 2004). In this case, when the
attacker starts to misbehave, all false identities will be
spread through the network, being part of several cer-
tificate chains. Note that, if there is any misbehaving
node in the chain, all other nodes of this chain might
obtain false authentications.
4 METRICS
Five metrics are used to evaluate PGP-Like:
• Certificate Exchange Convergence (CE);
• User Reacheability (UR);
• False Identity Confidence (FIC);
• Indirect Authentication of false identities (IA);
• Suspects Certificates by repository (SC).
The CE and UR metrics are used by (
ˇ
Capkun
et al., 2003) to evaluate PGP-Like in scenarios with
no attacks. The same metrics are used here to evalu-
ate PGP-Like against blackhole attacks. To evaluate
PGP-Like under Sybil attacks, we introduce the met-
rics: FIC, IA, and SC.
All these metrics consider: S as the set of system
nodes, |X| as the number of elements in set X and NC
as the subset of non-compromised nodes.
CE is the average percentage of certificates in the
local repositories of the nodes at the time t. It also
represents the convergence time, i.e., the time needed
by certificates to reach all nodes of the system. CE
can be defined as follows:
CE(t) =
∑
CE i(t)
|S|
∀i ∈ {S} in which (1)
CE i =
∑
|(K
a
K
b
) ∈ (G
i
∪ G
N
i
)|
∑
|(K
x
K
y
) ∈ G|
∀ a, b, x, y ∈ {S} (2)
UR is the average percentage of paths that node
i can find in its updated (G
i
) and non-updated (G
N
i
)
repositories at the time t. It represents the usefulness
of the certificate exchange mechanism for key authen-
tication. UR can be defined as follows:
UR(t) =
∑
UR i(t)
|S|
∀i ∈ {S} in which (3)
UR i =
∑
|(K
i
K
a
) ∈ (G
i
∪ G
N
i
)|
∑
|(K
i
K
x
) ∈ G|
∀a, x ∈ {S} (4)
FIC is the number of non-compromised nodes
that trust in a false identity. FIC can be defined as
follows:
FIC =
∑
FIC
i
|NC|
∀ i ∈ {NC} in which (5)
FIC
i
=

1 if ∃ m ∈ G
i
: m is a false identity
0 otherwise
(6)
IA is the ratio of non-compromised nodes (i) that
authenticate a false identity (m) using the merged up-
date repositories of i (G
i
) and m (G
m
). IA can be de-
fined as follows:
IA =
∑
IA
i
|NC|
∀ i ∈ {NC} in which (7)
IA
i
=

1 if ∃(K
i
K
m
) ∈ (G
i
∪ G
m
)
0 otherwise
(8)
SC is the fraction of certificates issued by a Sybil
node that can be found in the local repositories of the
non-compromised nodes. These certificates may or
may not be bound with a false identity. However, due
to the absence of a misbehavior detection mechanism,
these certificates are considered suspects. Let F be the
set of Sybil nodes, SC can be defined as follows:
SC =
∑
SC
i
|NC|
∀ i ∈ {NC} in which (9)
SC
i
=
∑
|(K
z
K
f
) ∈ G
i
|
|G
i
|
∀z ∈ {G
i
} and ∀ f ∈ {F}
(10)
QUANTIFYING MISBEHAVIOUR ATTACKS AGAINST THE SELF-ORGANIZED PUBLIC KEY MANAGEMENT
ON MANETS
131

5 EVALUATION RESULTS
The Network Simulator 2 (NS-2) was used to verify
the effectiveness of the PGP-Like scheme against two
misbehavior attacks, the Sybil and a variance of the
blackhole. The metrics CE(t) and UR(t) are used
to evaluate the effects of the blackhole attack, while
FIC, IA and SC
i
are used to evaluate the effects of the
Sybil.
The radio propagation model used is the two-ray
ground reflection and the link layer protocol is the
IEEE 802.11. Like (
ˇ
Capkun et al., 2003), simulations
use random certificate graphs, with 60 seconds cer-
tificate exchange interval. Also, certificate exchanges
are symmetrical and the network has no misbehav-
ior detection mechanism. Furthermore, for simplic-
ity, public and private keys are created by nodes only
during network formation. Certificates are also issued
during network formation: 600 trustful certificates are
issued between randomly selected pairs of nodes and
there is no certificate revocation. Note that these char-
acteristics were implemented in this way for simplic-
ity, not affecting the presented results. Other parame-
ters used for simulations are given in Table 1 and the
presented results are average of 35 simulations with
95% confidence interval.
Table 1: Simulations scenario parameters.
Parameter Used value
Network dimension 1000 x 1000 and
1500 x 300 meters
Power range 50 and 120 meters
Nodes 100 nodes
Mobility model random waypoint
Max. speed 5, 10 and 20 m/s
Max. pause time 20 seconds
Issued certificates 600 certificates
Exchange certificate interval 60 seconds
5.1 Blackhole Attack
The PGP-Like evaluation in the presence of black-
hole nodes considers 5%, 20%, 40%, 60% and 80%
of selfish nodes. These nodes only misbehave dur-
ing certificate exchanges, correctly working during all
network operations. They can even issue certificates.
Furthermore, selfish nodes request and accept certifi-
cates from other nodes, but they do not send certifi-
cates when they are requested to.
All simulation parameters are the same as the ones
in (
ˇ
Capkun et al., 2003): network lifetime is 1500 sec-
onds, 600 random certificates are issued at network
initialization. Results shown below are only for 20
m/s maximum speed, network dimension of 1000 x
1000 meters, transmission range of 120 meters and
100 nodes. Scenarios with 5 m/s and 10 m/s maxi-
mum speed, network dimension of 1500 x 300 me-
ters, transmission range of 50 meters and 50 nodes
were also evaluated, but these results are extremely
similar to the ones shown. Thus, they can be omitted
without losing generality.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
10 100 1000
CE(t)
Time (s)
Blackhole attacks
0%
5%
20%
40%
60%
80%
Figure 3: Convergence of Certificate Exchanges under
Blackhole Attacks.
Figures 3 and 4 illustrate the PGP-Like behavior in
presence of selfish nodes. Figure 3 shows the aver-
age percentage of certificates in the local repositories
of the nodes and, also, the convergence time. As ex-
pected, as the number of misbehavior nodes increases,
CE(t) decreases. Indeed, increasing the number of
misbehavior nodes, fewer nodes participate on certifi-
cate exchanges, affecting the amount of certificates in
the local repositories and the convergence time. The
impact of selfish nodes in CE(t) is very small with
5% of selfish nodes. Increasing the number of selfish
nodes to more than 40%, the impact in CE(t) also in-
creases. Also, it is possible to notice that up to 100
seconds, the presence of misbehaving nodes reduces
the effectiveness of the certificate exchange mecha-
nism in 15%, while after 1000 seconds, it is reduced
up to 70% with 80% of selfish nodes.
Figure 4 shows the effectivenessof PGP-Like, i.e.,
the user (node) reacheability using the local reposito-
ries of the nodes. Contradicting the theoretical assess-
ments (van der Merwe et al., 2007; Wu et al., 2007),
UR(t) has almost not been affected by the presence
of up to 60% of selfish nodes. After the certificate
convergence time, user reacheability is the same for
scenarios with 0%, 5%, 20%, 40% and 60% of selfish
nodes. In all these cases UR(t) is above 90%, show-
ing that, even though CE is compromised under 20%
of misbehavior nodes, the PGP-Like effectiveness can
be guaranteed up to 60% of malicious nodes.
SECRYPT 2008 - International Conference on Security and Cryptography
132

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
10 100 1000
UR(t)
Time (s)
Blackhole attacks
0%
5%
20%
40%
60%
80%
Figure 4: User Reacheability under Blackhole Attacks.
Note that CE(t) has a direct impact in the local cer-
tificate repositories (G
u
and G
N
u
) connectivity of all
nodes, i.e., the smaller the CE(t) is, the fewer paths
can be found in G
u
and G
N
u
. Thus, having almost
100% in UR(t), while having less than 70% in CE(t)
with 60% of misbehavior nodes means that almost
all nodes can build certificate chains to almost all
nodes, but these chains are limited to the 40% non-
compromised nodes. Moreover, if some of these cer-
tificates become invalid or these nodes leave the net-
work, UR(t) will decrease. Furthermore, under 80%
of selfish nodes, both the CE(t) and UR(t) are af-
fected. Therefore, such results must be discarded,
since the system might be compromised.
Nonetheless, misbehavior detection mechanisms
can minimize this problem. Such mechanisms might
be able to detect selfish nodes and block them. In this
way, PGP-Like could self-organize removing all cer-
tificates issued by these nodes from certificate chains.
5.2 Sybil Attacks
The PGP-Like evaluation in the presence of Sybil
nodes considers 5%, 10% and 20% of malicious
nodes. Two different Sybil nodes behavior were ana-
lyzed. In the first one, network lifetime is 3000 sec-
onds and Sybil nodes have a correct behavior dur-
ing the network initial phase, 1500 seconds. After
that, each one creates five false identities and issues
certificates to them, misbehaving for the remaining
1500 seconds. In the second one, network lifetime
is 1500 seconds and Sybil nodes misbehave since net-
work formation, issuing certificates that can be false
or not. All simulation parameters are the same as the
ones used in Blackhole attacks (Section 5.1).
Figures 5 and 6 show that false identities are dis-
seminated very fast. Figure 5 considers the first case,
in which Sybil nodes start acting after 1500 seconds,
i.e., they start creating false certificates after the net-
10
20
30
40
50
60
70
80
90
100
1500 1750 2000 2250 2500 2750 3000
FIC (%)
Time (s)
Sybil attacks
5% 10% 20%
Figure 5: Trustworthiness in False Identities Issued After
1500sec.
10
20
30
40
50
60
70
80
90
100
100 400 800
FIC (%)
Time (s)
Sybil attacks
5% 10% 20%
Figure 6: Trustworthiness in False Identities Issued at Net-
work Initial Phase.
work convergence time. It is possible to notice that
in less than 300 seconds all non-compromised nodes
have the false certificates in their local repositories.
Furthermore, this result is independent of the num-
ber of Sybil nodes. False identities are quickly propa-
gated by the certificate exchange mechanism and they
will reach all nodes in at most T
CE
.
Figure 6 considers the second case, in which Sybil
nodes start acting within the network. It indicates the
average percentage of possible false certificates. Note
that, even though false certificates are also spread over
the network, as in the previous case, the speed is much
smaller. This happens because Sybil nodes issue false
certificates before the convergence time of the “cor-
rect” certificates ends.
Note that, in both cases, false certificates will be
part of the non-updated repositories of the nodes,
meaning that nodes must verify them before using
them. However, a Sybil node can answer the verifica-
tion request and a non-compromised node might use
QUANTIFYING MISBEHAVIOUR ATTACKS AGAINST THE SELF-ORGANIZED PUBLIC KEY MANAGEMENT
ON MANETS
133

false certificates. If the PGP-Like misbehavior detect-
ing mechanism is used, all these certificates will prob-
ably be classified as un-specified. However, a Sybil
node can easily interact to change the certificate clas-
sification to non-conflicting.
If node u needs to authenticate a false identity m,
u merges its own updated repository with the updated
repository of node m (G
u
∪ G
m
), and tries to find a
certificate chain (a path) in the united repository. Fig-
ure 7 shows the percentage of indirect authentication
(IA), i.e., certificate chains in the united repository
that non-compromised nodes can achieve. In these
simulations, Sybil nodes create false identities at 100,
200, 300, 400 and 1500 seconds. Note that IA is inde-
pendent of the number of Sybil nodes and it increases
in time due to the certificate exchange mechanism.
10
20
30
40
50
60
70
80
90
100
5 10 20
IA (%)
Sybil nodes (%)
100 seg
200 seg
400 seg
800 seg
1500 seg
Figure 7: Indirect Authentication of False Identities.
Considering the local repositories of a node i, SC
i
is
the amount of certificate chains that have at least one
certificate issued by a Sybil node. Figure 8 shows the
number of certificate chains after 1500 seconds. Also,
it is possible to see that SC value increases within the
number of malicious nodes. These simulations con-
sider 5%, 10%, 20% and 40% of Sybil nodes in the
network. Note that, in scenarios with 5% of malicious
nodes, SC is almost 45%, while in scenarios with 40%
of malicious nodes, this value reaches 70%.
These results show that PGP-Like is completely
vulnerable to Sybil attacks, even with just a few Sybil
nodes in the network (5%). Thus, this work confirms
the theoretical assessments (Yi and Kravets, 2004; He
et al., 2007; Gouda and Jung, 2004) demonstrating
that Sybil attacks are great threats to PGP-Like. Fur-
thermore, this work quantifies the impact of Sybil
attacks against PGP-Like, demonstrating that its ef-
fectiveness is compromised independently from the
number of misbehaving nodes, enforcing the neces-
sity of security mechanisms to reduce the impact of
Sybil attacks.
10
20
30
40
50
60
70
80
90
100
5 10 20 40
SC (%)
Sybil Nodes (%)
Figure 8: Suspicious Certificate in Local Repositories.
6 CONCLUSIONS AND FUTURE
WORK
Among the key management schemes for MANETs,
PGP-Like is the main chaining-based key manage-
ment scheme. It is a self-organized public key man-
agement scheme, i.e., nodes create all pair-wised keys
themselves. They also issue certificates to other nodes
they trust. Two kinds of misbehavior attacks are con-
sidered to be great threats to PGP-Like, the imperson-
ating and the lack of cooperation attacks. Examples
of these attacks on MANETs are the Sybil and the
blackhole attacks, respectively. This work quantifies
the PGP-Like effectiveness under these two different
misbehavior attacks.
When submitted to the blackhole attack, PGP-
Like was able to maintain its effectiveness even in the
presence of more the 60% of selfish nodes, contradict-
ing previously theoretical assessments. However, the
convergence time was affect with only 20% of self-
ish nodes. Furthermore, as expected, as the number
of misbehavior nodes increases, the convergence time
decreases. Indeed, increasing the number of misbe-
havior nodes, fewer nodes participate on certificate
exchanges, directly affecting the amount of certifi-
cates in the repositories of the nodes and the conver-
gence time. In fact, even though PGP-Like is capable
of performing its basic operations, its functionality is
limited by the validity time of the non-compromised
certificates.
When submitted to the Sybil attack, PGP-Like
was completely vulnerable. Its functionality is af-
fected with just 5% of misbehaving nodes. Thus, this
work confirms the theoretical assessments demon-
strating that Sybil attacks are great threats to PGP-
Like. Furthermore, this work quantifies the impact
of Sybil attacks on PGP-Like, demonstrating that its
SECRYPT 2008 - International Conference on Security and Cryptography
134

effectiveness is compromised independently from the
number of misbehaving nodes, enforcing the neces-
sity of security mechanisms to reduce the impact of
Sybil attacks. Future work includes the development
of a key management scheme able to resist or even
reduce the impact of Sybil attacks.
REFERENCES
Agrawal, P., Ghosh, R. K., and Das, S. K. (2008). Co-
operative black and gray hole attacks in mobile
ad hoc networks. In Proc. of the 2nd Int. Conf. on
Ubiquitous Information Management and Com-
munication (ICUIMC ’08), pages 310–314.
Al-Shurman, M., Yoo, S.-M., and Park, S. (2004).
Black hole attack in mobile ad hoc networks. In
Proc. of the 42nd annual Southeastregional con-
ference (ACM-SE 42), pages 96–97.
Butty´an, L. and Hubaux, J.-P. (2003). Report on a
working session on security in wireless ad hoc
networks. Mobile Computing Communications
Review (SIGMOBILE), 7(1):74–94.
ˇ
Capkun, S., Butty´an, L., and Hubaux, J.-P. (2003).
Self-organized public-key management for mo-
bile ad hoc networks. IEEE Transactions on Mo-
bile Computing, 2(1):52–64.
ˇ
Capkun, S., Hubaux, J.-P., and Butty´an, L. (2006).
Mobility helps peer-to-peer security. IEEE
Transactions on Mobile Computing, 5(1):43–51.
Chlamtac, I., Conti, M., and Liu, J. J.-N. (2003). Mo-
bile ad hoc networking: imperatives and chal-
lenges. Ad Hoc Networks, 1(1):13–64.
Djenouri, D., Khelladi, L., and Badache, N. (2005).
A survey of security issues in mobile ad hoc and
sensor networks. IEEE Surveys and Tutorials,
7(4):2–28.
Douceur, J. R. (2001). The sybil attack. In Proc.
of the First International Workshop on Peer-to-
Peer Systems (IPTPS ’01), pages 251–260.
Engel, T., Fischer, D., Scherer, T., and Spiewak, D.
(2006). A survey on security challenges in next
generation mobile networks. In Proc. of The
Third Int. Conf. on Mobile Computing and Ubiq-
uitous Networking (ICMU’06).
Gouda, M. and Jung, E. (2004). Certificate disper-
sal in ad-hoc networks. In Proc. of the 24th
IEEE Int. Conf. on Distributed Computing Sys-
tems (ICDCS 04).
He, W., Huang, Y., Nahrstedt, K., and Lee, W. C.
(2007). Smock: A self-contained public key
management scheme for mission-critical wire-
less ad hoc networks. In Proc. of 5th IEEE Int.
Conf. on Pervasive Computing and Communica-
tions, pages 201–210.
Hegland, A. M., Winjum, E., Mjolsnes, S. F., Rong,
C., Kure, O., and Spilling, P. (2006). A survey
of key management in ad hoc networks. IEEE
Communications Surveys, 08(03):48–66.
Hubaux, J.-P., Butty´an, L., and
ˇ
Capkun, S. (2001).
The quest for security in mobile ad hoc net-
works. In Proc. of the 2nd ACM International
Symposium on Mobile Ad Hoc Networking &
computing (MobiHoc ’01), pages 146–155.
Michiardi, P. and Molva, R. (2003). Ad hoc networks
security. ST Journal of System Research, 4(1).
Ramaswamy, S., Fu, H., Sreekantaradhya, M., Dixon,
J., and Nygard, K. E. (2003). Prevention of coop-
erative black hole attack in wireless ad hoc net-
works. In Proc. of the Int. Conf. on Wireless Net-
works (ICWN ’03), pages 570–575.
van der Merwe, J., Dawoud, D., and McDonald, S.
(2007). A survey on peer-to-peer key manage-
ment for mobile ad hoc networks. ACM Com-
puting Survey, 39(1):1.
Wu, B., Wu, J., Fernandez, E. B., Ilyas, M., and
Magliveras, S. (2007). Secure and efficient
key management in mobile ad hoc networks.
Journal of Network and Computer Applications,
30(3):937–954.
Yi, S. and Kravets, R. (2004). Composite key man-
agement for ad hoc networks. In Proc. of The
First Annual Int. Conf. on Mobile and Ubiqui-
tous Systems: Networking and Services (Mobiq-
uitous’ 04), pages 52–61.
Zhang, Q., Wang, P., Reeves, D. S., and Ning, P.
(2005). Defending against sybil attacks in sensor
networks. In Proc. of the Second International
Workshop on Security in Distributed Computing
Systems (ICDCSW’05), pages 185–191.
Zimmermann, P. R. (1995). The official PGP user’s
guide. MIT Press, Cambridge, MA, USA.
QUANTIFYING MISBEHAVIOUR ATTACKS AGAINST THE SELF-ORGANIZED PUBLIC KEY MANAGEMENT
ON MANETS
135