Analysis of Attacks and Defense Mechanisms for QoS
Signaling Protocols in MANETs
Charikleia Zouridaki
1
, Marek Hejmo
1
, Brian L. Mark
1
, Roshan K. Thomas
2
and
Kris Gaj
1
1
ECE Dept., MS 1G5, George Mason University, 4400 University Drive,
Fairfax, VA 22030, U.S.A.
2
McAfee Research, McAfee Inc.,1145 Herndon Parkway, Suite 500,
Herndon, VA 20170, U.S.A.
Abstract. Supporting quality-of-service (QoS) in a mobile ad hoc network (MANET)
is a challenging task, particularly in the presence of malicious users. We present
a detailed analysis of attacks directed at disrupting QoS in MANETs. We in-
troduce a new class of attacks targeted at degrading QoS. We consider attacks
on both reservation-based and reservation-less QoS signaling protocols and dis-
cuss possible countermeasures. Finally, we identify and discuss the key issues in
achieving secure QoS provisioning in MANETs.
1 Introduction
Most of the literature related to security in MANETs [1],[2] to date has focused on the
problems of key management and secure routing. These problems do not address the is-
sue of protecting the network from attacks on QoS and denial-of-service. Cryptographic
techniques for ensuring the integrity and authenticity of routing messages can also be
applied to QoS signaling messages. However, cryptographic techniques by themselves
can only address a subset of the security problems that exist with current QoS signaling.
QoS provisioning introduces new vulnerabilities that are not addressed by secure
routing primitives. Attacks on routing are generally directed toward disrupting network
connectivity, whereas attacks targeted at QoS signaling need not affect connectivity. For
example, a route that is established by means of a secure routing protocol can still be
susceptible to attacks on QoS. If an attacker manages to compromise the key needed for
network authentication, it can become part of a “secure” route. Such a node may com-
ply with a secure routing protocol, but at the same time attack and exploit the signaling
protocol. Attacks on QoS can be carried out even by nodes that are not part of the route.
Securing QoS signaling is also challenging because some attacks against signaling may
be difficult to distinguish from legitimate network congestion conditions or loss of con-
nectivity. This paper aims to analyze a representative class of attacks targeted at QoS
This work was supported in part by the U.S. National Science Foundation under Grant CCR-
0209049.
Zouridaki C., Hejmo M., L. Mark B., K. Thomas and R. and Gaj K. (2005).
Analysis of Attacks and Defense Mechanisms for QoS Signaling Protocols in MANETs.
In Proceedings of the 4th International Workshop on Wireless Information Systems, pages 61-70
DOI: 10.5220/0002564400610070
Copyright
c
SciTePress
signaling in MANETs and identify the key elements required in a secure QoS signaling
scheme.
The remainder of the paper is organized as follows. Section 2 discusses the vulnera-
bilities of current QoS signaling protocols for MANETs. Section 3 provides an in-depth
analysis of attacks and defense mechanisms for QoS signaling in MANETs. Finally, the
paper is concluded in Section 4.
2 Vulnerabilities in QoS Signaling for MANETs
2.1 QoS Signaling for MANETs
The INSIGNIA protocol [3] is an example of a reservation-based QoS signaling proto-
col for MANETs. The SWAN protocol [4], is a reservation-less QoS signaling protocol
for MANETs. While INSIGNIA is based on the IntServ model of QoS in the Internet,
SWAN is more closely aligned with the Internet DiffServ model. Ad hoc SRRP [5] is
another QoS signaling protocol based on the IntServ paradigm. In this paper, we will
illustrate the attacks by applying them to INSIGNIA and SWAN, but the attacks have
wider applicability to QoS signaling protocols in general.
INSIGNIA In the INSIGNIA protocol, control information is piggybacked onto the
header of IP packets in order to reserve, renegotiate, and release resources for traffic
flows. When a source node wishes to reserve resources in a connection to a destination
node, the source sets a reservation (RES) mode bit in the header of an IP packet. The
header of a RES packet also indicates the level of QoS (“base QoS” or “enhanced QoS”)
and the minimum/maximum amount of bandwidth (e.g., 20/30 kbps) requested by the
flow. Upon reception of a RES packet, an intermediate node performs admission con-
trol, either accepting or denying the request, based on the availability of local resources.
The intermediate node modifies (if necessary) the packet header to indicate whether the
request is rejected, accepted at the minimum bandwidth, or accepted at the maximum
bandwidth and forwards the RES packet to the next node in the route to the destination.
If the request is accepted, the flow ID associated with the packet and the amount of
bandwidth reserved is recorded in a state table at the intermediate node. When the des-
tination receives the RES packet, it responds by sending a QoS report back to the source
node to indicate the amount of resources reserved for the flow, if any. Reservations are
made using “soft state, i.e., the reservation times out after a fixed time period, after
which resources are automatically released by all of the nodes on the route. In order to
maintain the reservation for longer durations, the source must periodically refresh the
reservation with new RES packets.
SWAN The SWAN protocol probes for available resources along a route between a
source and destination node without explicitly reserving resources for a given flow. The
source node sends a special control packet called bandwidth probe request BP R
eq
to
the destination node on a previously established route. The source node indicates its
bandwidth requirement in a field of the BP R
eq
packet called the bottleneck bandwidth
62
(BB) field. Upon receiving a BP R
eq
packet, an intermediate node on the route com-
pares the value in the BB field with the available bandwidth on its outgoing field and
overwrites the BB field if the available bandwidth value is smaller. When the destination
node receives the BP R
eq
packet, it reads the BB field and sends a bandwidth probe re-
ply packet BP R
ep
containing this value to the source node. After receiving the BP R
ep
packet, the source decides whether or not to establish a real-time flow given the rate in-
dicated in the BB field of the BP R
ep
packet. Since resources are not explicitly reserved
for each flow, network congestion may occur. Two mechanisms are provided in SWAN
to mitigate the onset of congestion: source-based regulation (SR) and network-based
regulation (NR) [4].
2.2 QoS Vulnerabilities of MANETs
The main vulnerabilities of MANETs with respect to QoS signaling are listed below:
1. Open network topology. In a MANET, the address and identity of a node are
independent of the node’s location. The open topology and overlaps in radio trans-
mission and reception ranges make it easier for attackers to overhear QoS requests
and control messages and to actively interfere with such messages. This makes the
signaling protocol vulnerable to attacks on confidentiality and availability.
2. Node mobility. In a fixed and wired network, the IP address of a host is considered
to be its identity and indicative of its location in a network topology. In a MANET
setting, it is difficult to trace and verify the legitimacy of QoS requests.
3. Intermittent connectivity. Due to intermittent connectivity, control messages may
be lost or protocol timing dependencies may be modulated. Such effects are difficult
to distinguish from real attacks.
4. Limited node capabilities. Typical nodes in a MANET have stringent resource
constraints such as limited energy, memory, and CPU cycles.
3 Analysis of Attacks on QoS Signaling
In this section, we analyze a representative class of attacks on QoS signaling. We as-
sume that a secure routing protocol is in place, i.e., a “secure” route between the source
and destination nodes has been determined prior to the initiation of QoS signaling. Our
analysis is given in terms of an attack template consisting of the following three com-
ponents:
– Vulnerability: the network state or property that the attacker exploits.
– Attack Step: the method by which the attacker carries out the attack, the position
of the attacker in the network, the amount of effort used by the attacker, etc.
– Effect: the observable effects and side-effects of the attack.
We also discuss possible countermeasures to each of the attacks.
3.1 Attacks on reservation-based QoS signaling
OVER-RESERVATION A greedy node can exploit the signaling protocol and reserve
more bandwidth for one of its real-time flows than what it actually needs to use. In
63
an extreme case, the greedy node could reserve bandwidth for non-existing flows in
order to perform a DoS attack or to ensure that its own real-time applications could be
supported in the near future.
Attack analysis: Vulnerability: (i) protocol cannot verify usage of reservations; (ii)
naive refreshment of reservations (e.g., INSIGNIA).
Step: attacker (i) acts as the source node; (ii) requests more bandwidth than it uses;
(iii) sends one data packet in the specified refresh-time interval to keep the reservation
refreshed.
– Effect: (i) bandwidth under-utilization; (ii) legitimate sessions are denied service.
Note that in the over-reservation attack, a link could be under-utilized even if no
additional real-time flows are initiated. Moreover, since the attacker does not need to
send many packets to launch this attack, the attack could be carried out over a long time
period.
Issues specific to MANETs: The wireless channel provides smaller capacities than
wired mediums. As a consequence, the over-reservation attack can create a DoS condi-
tion faster in MANETs than in wired networks. Furthermore, the use of straightforward
techniques for rate monitoring is often impractical due to the limited computational
power of the mobile nodes. A solution that does not overwhelm the node capabilities
should be sought.
Countermeasure: As a countermeasure, we propose that data rate monitoring and rate
adjustment should be performed by each node. Rate monitoring prevents the traffic
flows from under-utilizing their assigned rates. The data rate of an aggregate traffic
stream is measured and compared with the assigned rate recorded in the state table. If
the measured rate is lower than the assigned rate by a sufficient margin, the assigned
rate is decreased by a certain factor. This is the rate adjustment step.
To successfully apply these techniques in MANETs, they have to be scalable and
efficient. To avoid rate monitoring on a per-flow basis, each node could maintain state
for active aggregate traffic streams traversing the node per in-hop/out-hop pair. We de-
fine the “in-hop” node i to be the upstream neighbor node and the “out-hop” node j
to be the downstream neighbor node. An in-out stream through a node may consist of
many individual traffic flows. Rate monitoring becomes feasible when the traffic flows
through a node are aggregated and managed on an in-hop/out-hop basis such that the
node is not overwhelmed with too many rate monitoring computations. A protocol that
pursues this scheme is described in [6].
STATE TABLE STARVATION The state table starvation attack is another attack specific
to reservation-based signaling protocols, as such an attack is possible when the protocol
requires flow reservations, e.g., in INSIGNIA. It implies the reservation of state for
illegitimate flows and this leads to a state table exhaustion when the storage capacity of
a node is exceeded.
Attack analysis: Vulnerability: (i) node has limited memory and computational power;
(ii) reservations are made on a per flow basis; (iii) protocol cannot verify usage of reser-
vations.
Step: attacker (i) acts as the source of the data packets; (ii) requests bandwidth
for an illegitimate real-time flow.
64
– Effect: (i) state table is exhausted; (ii) legitimate sessions are denied service.
Issues specific to MANETs: Mobile devices are highly constrained in terms of memory
and can store only a limited amount of state information.
Countermeasure: An example of a countermeasure would be for each mobile node to
maintain a state table that grows as a function of the number of neighbor nodes, rather
than the number of traffic flows traversing the node. To avoid the storage of per-flow
state, each node could maintain state for each active aggregate traffic stream traversing
it on in-hop/out-hop basis, as discussed in the over-reservation attack. Thus, if a node
has N neighbors, the maximum number of in-out flows traversing the node is N (N 1).
3.2 General attacks on QoS signaling
OVER/UNDER-REPORTING OF AVAILABLE BANDWIDTH In this attack, a mali-
cious node on the path from the source to the destination node falsely represents the
available bandwidth on an outgoing link. For example, in SWAN, a malicious node on
a path could launch this attack by modifying the bottleneck bandwidth (BB) field of the
BP R
eq
message so as to falsely report the available bandwidth on its outgoing link.
Attack analysis: Vulnerability: (i) bandwidth availability is perceived differently by
different nodes due to the shared wireless medium; (ii) link capacities are not fixed due
to node mobility and wireless channel characteristics; (iii) protocol is unable to validate
the available bandwidth reported by an intermediate node.
– Step: attacker (i) acts as an intermediate node; (ii) falsely represents the available
bandwidth on its outgoing link in the BB field of a BP R
eq
packet.
– Effect: (i) source node sends at a rate that does not match the available bandwidth
on the path.
Fig.1. Over-reporting attack
Figure 1 illustrates an example of the over-reporting attack. Here, the source node S
requests 40 kbps for a real-time traffic flow by sending a BP R
eq
packet with a BB field
value of 40 kbps. Intermediate node I
1
performs an over-reporting attack by avoiding
the overwriting of the BB field of the BP R
eq
packet with a value of 10 kbps, even
though the available bandwidth on its outgoing link is only 10 kbps. Next, the BP R
eq
message is received by intermediate node I
2
, which does not overwrite the BB field,
since the available bandwidth on its outgoing link is greater than 40 kbps. The message
then reaches the destination node D, which creates a BP R
ep
message with a BB value
of 40 kbps. Upon receiving the BP R
ep
message, the source node proceeds to send its
65
data packets with a rate of 40 kbps, which will cause congestion on the link between I
1
and I
2
, as this can only support a rate of 10 kbps. The under-reporting attack is similar
except the node reports less bandwidth than what is available.
Issues specific to MANETs: Over-reporting/under-reporting attacks are more difficult
to detect and isolate in the wireless environment. The link capacities frequently change,
due to node mobility and wireless channel characteristics (interference, fading etc).
Moreover, several data link layer problems, such as the hidden/exposed terminal prob-
lems, exist. As a result, each mobile node perceives the wireless link capacity differently
and this can be exploited by malicious nodes to over-report/under-report their available
bandwidth. This is in contrast to wired networks, where the links are usually point-to-
point and the link capacities are fixed.
Countermeasure: Some MAC protocols for MANETs can reduce or virtually eliminate
the effects of the hidden/exposed node problems (cf. [7]). In this case, a node may be
able to estimate approximately the available bandwidth at a neighbor node by observing
the shared channel. Hence, the over-reporting/under-reporting attacks could be detected
by neighbor nodes, assuming they are not colluding as attackers.
Otherwise, the over-reporting attack could also be detected at the application layer.
In effect, the destination node could check the percentage of packets that were suc-
cessfully delivered. Isolating which node committed the over-reporting attack could be
done by triggering a search for a new route containing at least one link different from
the original path. If an over-reporting attack is detected along the new route, this infor-
mation could be provided to an IDS. The IDS could then attempt to isolate the attacking
node. Such an IDS would have to be “lightweight” enough to be implemented feasibly
on MANET devices. A conventional IDS would generally fail to detect this type of
attack and would be too computationally expensive to implement in a MANET.
The under-reporting attack cannot generally be detected in direct way. In this case,
the default action of the signaling protocol should again be to trigger a search for a new
route that has sufficient resources. The main effect of this attack is the extra overhead
required in searching for new routes. To mitigate this problem an IDS could log in-
formation on the amount of resource that is reported by nodes on different routes and
try to detect inconsistencies. The IDS could help to avoid choosing routes (if possible)
containing suspect nodes in the future.
QOS DEGRADATION QoS degradation represents a new class of attacks in QoS sig-
naling. It involves increase in the delay or jitter of the real-time packets to unacceptable
levels.
Attack analysis: – Vulnerability: (i) protocol does not verify QoS performance.
– Step: attacker (i) acts as an intermediate node; (ii) increases the delay or jitter of
the data packets to unacceptable levels.
– Effect: (i) QoS for a particular service is degraded; (ii) real-time session needs to
be re-initiated.
Increasing the delay or jitter of the real-time packets to unacceptable levels are
attacks specific to real-time flows. Conventional DoS mitigation techniques [8] cannot
recognize the increase on delay or jitter of the real time packets. Thus, the current
66
DDoS-aware IDS schemes cannot defend the network against QoS degradation attacks
in the Internet or wireless networks.
Issues specific to MANETs: QoS degradation attacks are difficult to distinguish from
impairments caused by the mobility of nodes or intermittent connectivity in the MANET.
Moreover, monitoring QoS is a particularly difficult task for mobile devices in MANETs
due to their limited capabilities.
Countermeasure: A QoS degradation attack could be detected at the application layer
by the destination node. As in the suggested countermeasure for the under-reporting/over-
reporting attack, detection of QoS degradation should trigger a search for a new route.
The presence of QoS degradation on the original route could be reported to an IDS,
which could help to avoid problematic routes in the future.
TIMING ATTACK The timing attack exploits the sequence in which signaling mes-
sages are sent or the timers defined by the protocol, with the objective of disturbing
the operation of the protocol. Both reservation-based or reservation-less signaling pro-
tocols can be susceptible to this type of attack. However, INSIGNIA in particular, does
not have easily exploitable timing dependencies and so is not susceptible to the timing
attack.
Attack analysis: Vulnerability: (i) protocol has timing dependencies; (ii) compliance
to protocol is not checked.
Step: attacker (i) acts as the source or destination node; (ii) exploits the timing
dependencies.
– Effect: (i) attacker gains access to the channel at the expense of legitimate flows.
Fig.2. Timing Attack 1.
SWAN implements regulate mechanisms [4] that can be easily exploited. As an
example, Figure 2 shows two sources S
1
and S
2
sending real-time data (at rates of
30 Kbps and 10 Kbps, respectively) to the same destination D. The link between the
intermediate nodes B and C experiences congestion since it has a capacity of only
30 Kbps. Node C is an attacker node that colludes with S
1
so as to give S
1
priority
over S
2
. The SWAN protocol requires that the intermediate node B set the ECN (Early
Congestion Notification) bit in the IP datagram of all the real-time flows traversing in
order to notify the destination D that congestion has occurred. The attack proceeds as
67
follows. Upon receiving this datagram, node D sends regulate messages to S
1
and S
2
.
However, attacker node C is on the path and chooses to delay the regulate message
destined for S
1
. As such, S
2
will attempt to re-initiate its session upon reception of
the regulate message and when its regulate-timer expires, whereas S
1
will continue
sending data. Consequently, S
2
will be required to probe the channel again to check
for the availability of resources, but will receive a BP R
ep
(Bandwidth Probe Reply)
message indicating that the channel is busy, as S
1
is still sending data. In summary,
by modifying the timing of the regulate messages, the attacker node C allows S
1
to
preempt S
2
and to maintain access to the channel. Clearly this attack can be used as the
basis for more sophisticated DoS attacks.
Issues specific to MANETs: The timing attack exploits the sequence in which signaling
messages are sent or the timers defined by the protocol. Timing attacks that exploit the
sequence in which signaling messages are sent, might not be recognized in MANETs,
as the intermittent connectivity and dynamic topology cannot guarantee that a message
will arrive at the destination or that it will reach the destination on time.
Countermeasure: The countermeasure here is to implement a QoS signaling scheme
that does not present time dependencies and one that does not employ timers to control
the protocol’s behavior. INSIGNIA, unlike SWAN, is not vulnerable to timing attacks.
FLOODING Neither reservation-based nor reservation-less signaling protocols are re-
sistant to flooding DoS attacks.
Attack analysis: Vulnerability: protocol (i) does not verify resource usage; (ii) does
not identify the source of flooding; (iii) does not take measures against flooding.
– Step: attacker (i) acts as the source node; (ii) floods the network with data traffic.
– Effect: (i) network is flooded; (ii) legitimate sessions are denied service..
Flooding the network with data traffic that consumes all of the available bandwidth
is a type of a bandwidth depletion attack. However, in INSIGNIA other types of DoS
attacks such as over-reservation and state-table starvation attacks can be launched with
much less effort.
Issues specific to MANETs: One technique to mitigate flooding is to trace back the
attacker and cut off the attack traffic at the source. However, it is much more challenging
to trace back an attacker in MANETs than in the wired environment.
Countermeasure: As a countermeasure, the traffic flows should be policed so that they
do not exceed their reserved rates. In order to avoid policing on a per individual flow
basis, aggregate traffic streams can be policed on an in-hop/out-hop basis as discussed
above (cf. over-reservation attack). If an individual flow transmits above its assigned
rate, it may experience traffic policing by at least one of the intermediate nodes on
the path. Such a distributed traffic policing scheme minimizes the amount of state and
processing required in each node of the MANET.
REPLAY ATTACK Any protocol that allows the exchange of unauthenticated informa-
tion is vulnerable to modification and replay.
Attack analysis: – Vulnerability: (i) protocol does not protect the integrity of signaling
information; (ii) protocol cannot distinguish a replay from an authentic message; (iii)
open topology.
68
Step: attacker (i) duplicates/modifies signaling information; (ii) forwards modi-
fied packet to the next hop.
– Effect: (i) resources are wasted by illegitimate packets; (ii) legitimate packets are
denied service.
Fig.3. Replay Attack
The replay attack can be performed by a compromised node on the route to the
destination by duplicating and modifying the information in a signaling message. In
the example shown in Figure 3, the malicious node M receives the control packet sent
by the source S
1
to the intermediate node A. Node M modifies specific fields in the
control packet and forwards the packet to the next hop. If the modified packet reaches
the destination first, it will be accepted while the original packet that traverses though
the route will be disregarded as a duplicate packet.
As an example of a specific replay attack, node M might choose to lower the Bot-
tleneck Bandwidth (BB) field of the control packet to falsely indicate that there is not
enough bandwidth for the establishment of the requested real-time session, in order to
deny S
1
access to the channel. If the routing protocol requires hop-by-hop authentica-
tion, node A will authenticate the packet before forwarding it. Since node M does not
need to authenticate the packet, it may be able to forward the replayed packet to the
destination before node A.
Issues specific to MANETs: Since the wireless channel is a broadcast medium, each
mobile node hears the transmission of every node in its radio transmission range. In
contrast, Internet switches and routers forward packets to destination nodes through
designated ports. A node in a wired network can generally only see the packets destined
to it.
Countermeasure: As a countermeasure, the number of replayed packets that are processed
should be limited over a time interval. If the number of packets with ID i, received at the
destination within a predefined time interval, is smaller than a threshold value, all the re-
ceived packets should be processed. Otherwise, only the first received packet should be
processed. This limits the number of packets of the same ID that can flood the network.
69
4 Discussion and Conclusion
We have shown that current proposals for QoS signaling in MANETs are highly suscep-
tible to a number of powerful attacks, even when a secure routing protocol is in place.
Our investigations suggest that a complete solution to secure QoS signaling solution for
MANET should incorporate the following elements: (1) intelligent traffic management,
(2) lightweight intrusion detection, and (3) efficient cryptographic primitives.
As discussed, a large class of QoS-based attacks in MANETs can be mitigated via
distributed traffic management (cf. [6]). However, some attacks are difficult to mitigate
without some means of identifying which nodes have been compromised. Conventional
intrusion detection systems are generally impractical for MANETs (cf. [9], [10]), but
distributed trust establishment schemes (cf. [11]) could provide sufficient information
for nodes to avoid certain types of QoS signaling attacks. While we have not focused on
the application of cryptographic primitives to QoS signaling, some form of lightweight
scheme is necessary to authenticate the signaling control information.
In ongoing work, we are further developing the DRQoS scheme presented in [6]. We
are also investigating computationally lightweight schemes to establish trust measures
for MANETs that could be used to make QoS signaling more secure. Our ultimate aim
is to design a flexible and secure QoS signaling protocol that can successfully resist
attacks which exploit the inherent vulnerabilities of MANETs.
References
1. Zhou, L., Haas, Z.J.: Securing Ad Hoc Networks. In: IEEE Network Special Issue on Net-
work Security. Vol. 13. (1999) 24–30
2. Hu, Y.C., Perrig, A., Johnson, D.B.: Ariadne: A Secure On Demand Routing Protocol for Ad
hoc Networks. In: Proc. ACM MobiCom ’02. (2002) 12–23
3. Lee, S.B., Ahn, G., Zhang, X., Campbell, A.T.: INSIGNIA: An IP Based Quality of Service
Framework for Mobile Ad Hoc Networks, In: Journal of Parallel and Distributed Computing.
Vol. 60. (2000) 374–406
4. Veres, A., Campbell, A.T., Barry, M., Sun, L.H.: Supporting Service Differentiation in Wire-
less Packet Networks Using Distributed Control (SWAN). In: IEEE Journal on Selected Ar-
eas in Communications. Vol. 19, (2001) 2094-2104
5. Yeh, C.H., Mouftah, H.T., Hassanein, H.: Signaling and QoS Guarantees in Mobile Ad hoc
Networks. In: Proc. IEEE ICC. (2002) 3284-3290
6. Hejmo, M., Mark, B.L., Zouridaki, C., Thomas, R.K.: Denial-of-Service Resistant Quality-
of-Service Signaling Protocol for Mobile Ad hoc Networks. In: Proc. ACM SASN Work-
shop. (2004) 23-28
7. Haas, Z.J., Deng, J.: Dual Busy Tone Multiple Access (DBTMA): A Multiple Access Control
Scheme for Ad Hoc Networks. In: IEEE Trans. on Comm. Vol. 50. (2002) 975–985
8. Mirkovic, J., Reiher, P.: A Taxonomy of DDoS Attacks and Defense Mechanisms. In: ACM
SIGCOMM Computer Communications Review. Vol. 34, (2004) 39-54
9. Huang, Y.A., Lee, W.: A Cooperative Intrusion Detection System for Ad hoc Networks. In:
Proc. ACM SASN Workshop. (2003) 135–147
10. Tseng C.Y. et al: A Specification-Based Intrusion Detection System for AODV. In: Proc.
ACM SASN Workshop (2003) 125-134
11. Eschenauer, L., Gligor, V.D., Baras, J.: On Trust Establishment in Mobile Ad-hoc Networks.
In: Proc. Security Protocols Workshop. Vol. 2845 (2002) 47-66
70