Trust: An Approach for Securing Mobile ad hoc
Network
Chung Tien NGUYEN and Olivier CAMP
Departement of Computer Science, ESEO
ANGERS, France
Abstract. When functionning in the ad hoc mode, wireless networks do not rely
on a predefined infrastructure for achieving the basic network functionalities.
Hosts of such networks need to count on one another to keep in contact with
the network and carry out services such as routing, security, auto-configuration,...
Network services and, in particular, security thus strongly depend on the way
the nodes find the correct partners with which they can cooperate efficiently. As
consequence, it seems important for ad hoc networks to provide a representation
of trust together with a mechanism to evaluate it. In this paper, we present ad
hoc networks, and show how trust is fundamental in the existing propositions to
improve their security. After identifying the characteristics of existing trust mod-
els, we focus on those that should be implemented in a trust model for ad hoc
networks.
1 Introduction
Wireless technologies offer, today, new possibilities in the fields of telecommunica-
tions. The development of wireless communication allows manipulation of information
through mobile calculating units such as laptop computers, personal organisers, mobile
phones, sensors,... These units should be capable of accessing the network through a
wireless communication interface and are able to roam freely.
Such mobile environments, which are composed of mobile units interconnected by
radio links, allow a very flexible way of implementing communicating applications in
various fields. In particular, they allow the establishment of networks in sites where it is
difficult, or even impossible, for an infrastructure to be installed - eg; construction sites,
mobile laboratories, search and rescue operations, battlefields, ...
Wireless mobile networks can be divided into two classes: infrastructured mobile
networks (i.e cellular network,...), and mobile ad hoc networks which are self organised
and do not need an infrastructure. There are no dedicated routers, servers, access points
or cables in these networks and theirs entities can join or leave the networks at any time.
The security provided by ad hoc networks should be equivalent to that expected
from traditional infrastructured wired networks. However, due to the lack of infrastruc-
ture and entry point, it is difficult to transpose existing solutions to this type of context.
In such a complex environment, it is difficult for an entity to determine which enti-
ties are malicious and which are not. This is a very important point since entities in it
need to communicate and cooperate in order to achieve basic network services. Several
Tien NGUYEN C. and CAMP O. (2005).
Trust: An Approach for Securing Mobile ad hoc Network.
In Proceedings of the 4th International Workshop on Wireless Information Systems, pages 92-103
DOI: 10.5220/0002574000920103
Copyright
c
 SciTePress

techniques can be used to identify entities that are authorized to join the network but
such restrictions tend to degrade the network’s efficiency. It is necessary for the entities
to be able to determine the trust they can have in each other and, based on this trust,
determine with which entities they can cooperate. A trust model allowing an entity to
determine to what extend it can trust the others seems to be necessary to improve the
security of ad hoc network.
Our goal is to develop a fully decentralised trust model, suitable for ad hoc net-
works, a dynamic and completely distributed environment.
The rest of the paper is structured as follows: in section 2 we present ad hoc net-
works, existing propositions to provide security to these networks and show that trust
is a fundamental issue in most of them. Section 3 does a brief survey of existing trust
models and introduces the key notions of reputation, recommendation and reciprocity.
In section 4, we go through some important characteristics of existing trust models and
identify those that seem important for ad hoc networks. Section 5 concludes the paper
and discusses potential directions for future works.
2 Ad hoc networks
A mobile ad hoc network (MANET) is a wireless network made up of autonomous and
mobile nodes. It does not need a predefined infrastructure to exist. It is created spon-
taneously by a temporary association of mobiles, which communicate through radio
links. If the diameter of the network exceeds the radio transmission range, nodes have
to participate in a routing protocol that considers the dynamic aspects of the network.
This type of network has the following characteristics:
– A dynamic topology that evolves rapidly in time according to the movement of mo-
biles, their emission power, and the characteristics of the communication channel.
– Unreliable connections with variable flow and limited bandwidth.
– Battery powered nodes that thus have a limited lifetime.
– An open network without any entry point on which administrative services can be
installed
Implementation of security solutions in such an environment is a hot and challeng-
ing issue.
Firstly, MANETs are open. Since they have no clear entry point, a node can integrate
the network by simply entering the transmission zone of another node. There is thus a
probability of accepting a potential attacker into the network. Therefore, we should not
only consider malicious attacks from outside the network, but also those launched from
within the network by compromised nodes.
Moreover, the performances of MANETs do not allow for all security services to be
implemented. Wireless connections in a MANET make it quite easy to carry out pas-
sive eavesdropping, active impersonation, message replay, and distortion. When within
radio range of a node, an attacker can eavesdrop the messages it transmits (violate their
confidentiality), delete these messages, and modify them (violate the availability and
integrity).
93

Furthermore, MANETs are dynamic because their topology may change frequently
and nodes may often join and leave the network. Security solutions based on a static
configuration of the network would thus not be sufficient.
Finally, a MANET can become enormous and contain hundreds or, even thousands,
of nodes. Security mechanisms should be scalable to handle such large networks.
Many security solutions have been proposed to implement security in MANETs.
In [1], the authors propose a solution based on group management. The whole net-
work is a hierarchy of special purpose groups and sub-groups defined in accordance
to the application context. The hierarchy signifies that a group can comprise several
sub-groups and each sub-group can comprise several sub-sub-groups. A management
framework is provided for establishing and managing this hierarchy. This framework
supports control on group composition and allows for basic operation such as group
joins and leaves. A manager administrates each group and sub-group. It is responsible
for creating and maintaining group and for receiving and processing group operations
sent by users. The manager is selected after a mechanism described in [2], this mecha-
nism allows automatically selecting a manager by taking into account the status of the
network and the capacity of the node. Mobile code is used to manages group and to
determine if a node can become a manager.
A solution, which is considered as a method to strengthen the security in MANETs,
is intrusion detection. Intrusion detection systems (IDS) allow detecting violations aga-
inst the security policy. In a MANET, it is difficult to analyze network activities globally.
Each node only possesses a limited vision of the network’s activities. This limit depends
on the characteristics of stations and is an important constraint for intrusion detection
algorithm. Each station thus needs its own intrusion detection system, and makes it
participate in the network’s global intrusion detection mechanism.
[3] and [4] propose this kind of IDS. [3] uses independent agents on each station
to locally detect intrusions. When a local anomaly is detected or evidence is not clear,
the agents participate in a global detection. [4] proposes an architecture for IDS in
which information is collected and exchanged using mobile agents. To determine if an
intrusion is occurring, a node uses local information and more information gathered
from remote nodes by the agents. Information collected remotely will only be usable if
it can be trusted, i.e if the nodes it was obtained from are trusted by the gathering agent.
[5] proposes another solution for security: secure routing protocol. The authors add
the notion of trust to the routing protocol AODV (Ad-hoc On demand Distance Vector
routing). In this protocol, routing information is encrypted so that a malicious node can-
not know who the sender is and nodes included in the route are authenticated. Encryp-
tion levels are selected in accordance with the trust levels existing between successive
nodes and the level of security required by the application, needing the route. However,
[5] does not specify how the trust level between nodes is determined.
In [6], Prashant Dewan and Partha Dasgupta make use of the concept of reputation
in the routing protocol of MANETs. Reputation of an entity is determined through
its past behavior. It is used to calculate probability that transactions between nodes
are satisfied. The reputation of an entity is increased when it successfully transfers
data packets and decreased when it does not. Each entity keeps information about the
94

reputation of nodes it knows and transfers this information, in recommendation, to other
entities. The mechanism supposes that entities do not falsify their reputation.
Trust and reputation seem to be important factors in the field of security in ad hoc
network and research in this field is very active. In a network in which many different
kinds of nodes and users coexist, the notions of trust and reputation are necessary but
difficult to realize.
3 Trust and reputation
3.1 Definition
Trust is a decisive factor in collective performance, particularly in virtual communities.
Literature on this subject is plentiful and definitions are various, for example:
– Trust proceeds from reasoning, the ratio effort-benefit of an individual action within
a collective.
– Trust is based on a nomination such as a label, a certificate. For example, we are
able to confide our health and our life to an unknown doctor just because he ob-
tained a national diploma, which we do not even care to verify. In electronic com-
merce, we use credit card numbers to buy merchandises on web sites, which possess
well-known digital certificates.
– Trust comes from intuition, from belief, which do not suppose a true deliberation.
This trust is emotional, aesthetic and irrational.
– Trust is based on a sort of ”engagement to respect the norm” that comes from a
rule of reciprocal duty. This trust is defined as ”an expectation on the motivation of
the other to behave in accordance with whatever is predicted in a given situation”.
This definition considers individuals as rational and foreseeable actors, and their
rationality is strengthened by their correct choices and their acts [7].
The theory of rationality supposes that individual choices rely on utilitarian reasoning:
– If I prefer A to B and B to C, then I prefer A to C
– All decisions are based on cost-benefit ratio, or on a risk analysis.
Certain try to put trust in relation with other concepts such as cooperation, recommen-
dation,
Gambetta, in [8], establishes a connection between trust and cooperation and a cer-
tain degree of trust is required to realize cooperation. If trust is unilateral, the coopera-
tive task cannot be realized. Thus, the higher degree of trust, the higher the possibility
of realizing cooperation.
Recommendation plays an important role in the field of trust since it is impossible to
trust everybody. When someone does not know whether he can put his trust in someone
or not, she tends to ask a trusted third person. Suppose that A wants to know how he
can place his trust in unknown X: if A can get a recommendation concerning X from
a trusted entity B, the trust granted to X is function of this recommendation and of the
trust granted to B. If B has no information concerning X, she will ask other entities. In
95

general, the longer the chain of recommendation, the lower the possibility of granting
trust to the concerned person.
Patha Dasgupta gives another view of trust in [9]. According to him, though a mea-
sure of trust does not exist, it is possible to calculate the level of trust through inter-
mediary. Dasgupta considers trust as knowledge or information. In his opinion, trust
is expectation on activities of an entity when it reacts in a given context. Dasgupta
concludes that trust is based on reputation, which is constructed through behaviors in
known circumstances.
3.2 Direct trust and recommendation
In [10], Beth, Borcherding and Klein formally present trust relationship. They consider
a system consisting of entities, which communicate via links. Some trusted entities,
called Authentication Server (AS), play the role of authenticating agents for other en-
tities. That means, when one entity wants to obtain information berfore engaging in a
new experiences with others, it needs to ask an AS. If this AS cannot do that, it will ask
another AS and become a mediator of the experience. The degree of trust is based on
the number of positive end negative experiences.
[10] defines six classes of trust: key generation, identification, secret keeping, non-
interference, clock synchronization and performance of algorithmic steps. Each class
can have two types of trust: direct trust and recommendation trust. Direct trust is granted
directly to the other entity while recommendation trust is based on the recommendation
of a third entity.
In [10], direct trust is defined as follows:
A trust
seq
x
B value V
A direct trust relationship exists if all A’s experiences with B with regard to trust
class x are positive. Seq is a sequence of entities that mediated the experiences between
A and B. The trust value, V, is an estimation of the probability that B behaves well. This
is base on the number of positive experiences A has had with B.
Recommendation trust is defined as follows:
A trust
seq
x
B when.path S
p
when.target S
t
value V
Recommendation trust exists if A accepts reports from B about experiences with
third parties with regard to trust class x. The third parties are restricted to the entities in
S
t
(the target constraint set) and the mediators of the experiences are restricted to S
p
(the path constraint set).
The example of derivation of a trust relationship with a recommendation in figure 1
is given in [10]. With the help of some rules, a new trust relationship can be established
from a set of initial relationships.
Based on existing trust relationships, shown in figure 1(a), the new trust relationship
between A and C and between A and D can be derived. The derivation can be presented
by the following equations:
96

Fig.1. Derivation of trust
Derived trust between A and C
V
1
⊙ V
2
= 1 − (1 − V
2
)
V
1
Recommendation trust between A and D
V
1
· V
3
= simple multiplication between V
1
and V
3
This multiplication shows that the value of derived recommendation trust decreases
when the recommendation chain grows.
3.3 Relationship between trust, reputation and reciprocity
[11] proposes a model for calculating trust and reputation for E-Business. This model
defines three concepts: trust, reputation and reciprocity. Trust is the subjective ex-
pectancy of an individual about the future behavior of others, based on their previous re-
lations. The reputation of an entity is the perception of individuals, and created through
its activities in the past. Each individual shows her reciprocity if she responds correla-
tively to what others give to her - i.e; she collaborates with people who collaborate with
her and has a negative behavior with those who do not.
With this model, the author expects individuals to trust a person, who has a high
reputation and to be cautious with people, who don’t. When an individual frequently
shows her reciprocity, she can incite others to increase her reputation.
Suppose that an individual a
j
wants to estimate the reputation of individual a
i
to
establish cooperation. The set A of N individuals, which a
j
asks for information about
a
i
, is called the ”embedded social network”: A = {a
1
, a
2
,...,a
N
}
Reputation is diffused so that individuals, who cooperate well, are rewarded. The
following relations are expected:
– Augmentation of reputation of a
i
in A increases the trust, which is granted to a
i
by
A.
97

– Augmentation of trust in a
i
increases the possibility that a
j
responds positively to
actions requested by a
i
– Augmentation of the reciprocity between a
i
and other members of A increases the
reputation of a
i
in A
However, the activity space of this model is binary: Two nodes either cooperate or they
do not.
4 A Trust model for ad hoc mobile networks
MANETs evolve in a completely open environment. Many nodes can join the network
for different purposes: researchers want to discus with each others, professors want to
exchange their opinions about courses or exams, students want to do exercises together,
some people, if possible, just want to know what others are doing. When belonging to
a MANET, a node usually goes in communication with a large number of other nodes.
However, hosts do not necessarily trust all the others, sometimes simply because they
do not know anything about each other. They need to be able to choose good nodes,
with which they can cooperate to complete their task. This choice depends on the trust
between them; and nodes should be able to evaluate the trust they have in each other.
A model allowing the expression of trust among nodes seem necessary for MANETs.
Existing models are applied in different security domains: PGP (Pretty Good Pri-
vacy), Maurer-D and SPKI by public key infrastructures, X509 by certification au-
thorities, Poblano in peer-to-peer environment,... These models possess characteristics
specifically adapted to their domain. They can be classified as follows:
– The types of trust represented in the model.
– The representation of trust.
– The possible evolutions of trust.
– The centralization of the model.
4.1 Types of trust
The trust granted to an entity can be:
– Direct trust: based on historic relation with the entity
– Recommendation trust: based on the recommendation of a third party concerning
the entity’s behaviors.
Direct trust is essential to a majority of models. Based on its experiences with other
entities, an entity can decide to cooperate with them. But direct trust is not sufficient.
MANETs are a complex environments and evolve constantly, their members change
and the number of member may be large. An entity should therefore be able to place
its trust in others through one or several recommendations obtained from third parties.
The problem of recommendation trust is more complex than that of direct trust. In the
simplest case, trust in an entity can be assumed if there is a recommendation from
somebody in whom we have direct trust. In more complex cases, trust can be calcu-
lated according to the experiences for which the recommendation holds. That helps in
determining more precisely the range of trust in the recommended entity.
98

4.2 Representation of the trust
In certain trust models, reputation is used to determine trust between entities. Supposing
that an entity A trusts an entity B for achieving a particular service, data transfers, for
example. In this case, A can say that B has a reputation of being reliable in transferring
data. Reputation can be deduced from behaviors of an entity in past occasions.
In order to represent reputation or trust, a model can use a binary representation or
a degreed representation. The first one allows to represent only two states: complete
trust or no trust at all. This representation is simple and avoids ambiguous interpreta-
tions. It is used in many implementations: X509, SPKI,... However, it is sometimes too
restrictive. Since the trust accorded to an entity may not be complete but sufficient for
cooperation.
In degreed trust, there are more than two states of trust. The degrees of trust can be
discrete or continuous.
Considering the presentations of trust in Maurer-D [12] or PGP [13], trust is rep-
resented by discrete degrees. PGP uses public key cryptography to encrypt electronic
mails and uses a graph based approach of trust to certify public keys. Trust in PGP is
certainty that a public key belongs to a specific individual. There is no certificate au-
thority that is trusted totally and signs all public keys. Instead, individuals sign public
key for others and progressively establish a web of public key, which is interconnected
by signatures. An individual P trusts the public key of an individual Q (P does not know
Q) if it is signed by individuals (introducers) which are trusted by P. The level of trust
granted to an entity is function of the number of recommendation it has and who we got
them from (their introducers). Various degrees of trust can be granted to a public key.
They are categorized in PGP as follows:
– Undefined: we cannot say whether this public key is valid or not.
– Marginal: this public key may be valid be we cannot be too sure.
– Complete: we can be wholly confident that this public key is valid.
– Ultimate: the public key owners construct locally.
Levels of trust are granted to introducers:
– Full: recommendations from this introducer are always trusted.
– Marginal: recommendations from this introducer are only trusted partially.
– Untrustworthy: recommendations from this introducer are not trusted.
– Unknown: this introducer is not known.
Some models adopt a finer representation of trust and represent it by a continuously
evolving value. This value represents the probability that the concerned entity collab-
orates correctly. Trust may be the result of a reasoning based on objective events or
subjective beliefs: we talk about subjective logic and objective logic.
In the case of MANETs, a degreed representation seems more suitable. Indeed,
cooperation in MANETs is various, it is realized for different purposes and does not
necessarily need to use the highest levels of trust at all times. Using a degreed represen-
tation, entities have more chances to cooperate and complete their tasks. Furthermore,
with this representation, the security policy of MANETs can be specified more precisely
by fixing necessary thresholds of trust to carry out different types of cooperation.
99

4.3 Evolution of trust
Evolution of trust is made up of two main phases:
– The initial trust formation phase: this phase allows initialization of the trust granted
to an entity
– The revision phase: this phase updates the trust in others.
The initial trust formationphase In this phase, an entity initializes its trust in others. It
can be described by user defined assertion or automatically computed by an algorithm.
PGP is an example of manual description of initial trust. The user specifies the
degree of trust he has in an introducer. This trust determines how recommendations
from this introducer should be considered.
Concerning the automatic formation, it can be constrained or not.
Without constraints, trust is safely based on recommendation. For example, in [12],
an entity places its trust in others when a trust derivation rule is satisfied. The general
rule is that, for all i≥1, P trusts Q level i, if R, whom P trusts with level i + 1, makes a
recommendation for Q. The formal description is as follows:
∀ Q, R, i ≥ 1 : Aut
P,R
, T rust
P,R,i+1
, Rec
R,Q,i
⊢ T rust
P,Q,i
where Aut is the statement for P’s belief in the authenticity of R’s public key, Rec
indicate the recommendation, and Trust is the statement indicating trust.
With constraints, the deduction allowing to grant trust to an entity is restricted by
constraints. For instance, [14] defines the Time-To-Live constraint as the maximum
length of the recommendation chain.
In X509, in the case of certification authority system, the constraints are the certifi-
cates’ extension fields, which allow to limit the number of accepted certificates.
The revision phase Trust in an entity is not constant, it evolves in its life. The variation
of the degree of trust depends on many factors: experiences of the entity, the context in
which it evolves. Trust evolves when:
– The entity realises that the value of trust it possesses is inaccurate
– Another entity seems to be better than the presently most trusted.
Trust among entities can be function of experiences among them. Reagle Jr. [15] pro-
poses a model to manage reputation between sellers and buyers in an electronic com-
merce system. The evolution of reputation is based on the most recent experience be-
tween agents. It is simply a change from trust to distrust and leads to an interruption of
interactions with the suspected agent.
BBK [10] uses positive and negative experiences to make trust evolve. Direct trust
exists if there are no negative experiences with the entity in question.
These models use a simple variable to represent the value of trust in order to save
memory. This is not sufficient for determining the coherence between the most recent
100

experience with the entity and its activities in the past. Jøsang [16] proposes an ap-
proach based on knowledge to represent more information. This approach needs more
resources.
The information used and the way it is represented depend on the application of the
trust model and on available resources. In fact, the more effective the model, the more
resources it consumes. A balance between these factors needs to be found.
4.4 Centralization of trust model
Recommendations may come from different sources. In some cases, recommendation
comes from a central network entity (a user, an organization,...) which all members
trust: X509 is an examples. All members trust the Certificate Authorities’ (CA) public
key and only recommendations originating from the CA are accepted. In such a central-
ized model, trust in entities depends on a limited number of network entities: the CAs.
If the appropriate CAs are not available, trust in some entities can not be calculated. In
other cases, network entities can use recommendations from any entities they trust. This
is the case in PGP’s web of trust. A member in system can trust and choose anyone as
a introducer. Since then, he will believe in recommendations of this introducer. That is
suitable for activities in evolutionary environment. The lack of central point in ad hoc
network leads to the need of such a distributed trust system, in which entities can get
recommendation from several sources. A trust model for ad hoc network should thus
rely on a totally decentralised model.
Table 1. Characteristics of existing trust models
Type Representation Evolution
Model Logic Initial formation Centralized
Direct Recom. binary degree Subj. Obj. no contrs. with contrs.
BBK[10] × × × × × × ×
Jøsang [16] × × × × × × ×
Marsh[17] × × × × × ×
Maurer-D[12] × × × × × ×
PGP[13] × × × × ×
SPKI [18] × × × × × × ×
X509 [19] × × × × × ×
Sierra [20] × × × × × ×
Poblano [21] × × × × ×
Model proposed × × × × × × ×
The efficiency and the application context of a trust model depend on its charac-
teristics. In table 1, we summarize the characteristics proposed by existing models and
propose a list of characteristics that should, from our point of view, be implemented by
a trust model for MANETs.
101

5 Conclusion and future works
In this paper, we have presented MANETs, and certain aspects concerning their secu-
rity. We have also shown how trust concept is necessary for improving the security in
MANET, which characteristics existing trust models have. Finally, we have identified
aspects, which, in our opinion, should be implemented in a trust model wireless for
MANETs. This model should have the following characteristics:
– Model both direct trust and recommendation trust.
– Use a degreed representation of trust. A binary representation is insufficient for the
complex environment of MANETs.
– The initial trust formation should be automatic and manual. Trust must be evolu-
tionary. This requires knowledge about objective logic and subjective logic.
– Support a distributed system in which entities can get recommendations from dif-
ferent sources.
This work is only the preliminary phase in the establishment of a trust model adapted
to the nature of MANETs. One of the factors we will particularly focus on, will be the
decentralized nature of the model. In the next step of our work, we plan to fully specify
the model and see if it can be adapted to deal with other types of distributed applications
such as multi agent systems in which trust is also an important issues.
Acknowledgement
The work of Mr. Chung Tien NGUYEN is supported by a scholarship from Conceil
G
´
en
´
eral du Maine et Loire, France.
References
1. Meissner, A., Musunoori, S.B.: Group integrity management support for mobile ad-hoc
communities. In: Middleware Workshops. (2003) 53–59
2. Shen, C.C., Jaikaeo, C., Srisathapornphat, C., Huang, Z.: The guerrilla management archi-
tecture for ad hoc networks. Proceedings of IEEE MILCOM, Anaheim, California, USA
(2002)
3. Zhang, Y., Lee, W.: Intrusion detection in wireless ad-hoc networks. In: Proceedings of the
6th annual international conference on Mobile computing and networking. (2000) 275–283
4. Percher, J.M.: An Intrusion Detection Model for Wireless Ad Hoc Networks (in French).
PhD thesis, Universit de Versailles (2004)
5. Nekkanti, R.K., wei Lee, C.: Trust based adaptive on demand ad hoc routing protocol. In:
Proceedings of the 42nd annual Southeast regional conference. (2004)
6. Dewan, P., Dasgupta, P.: Trusting routers and relays in ad hoc networks. 2003 International
Conference on Parallel Processing Workshops (2003) 351
7. Prax, J.Y.: The role of trust in collective performance. In: Manual of Knowledge Manage-
ment - A Second Generation Approach. (2003)
8. Gambetta, D.: Can We Trust Trust? In: Trust: Making and Breaking Cooperative Relations.
Department of Sociology, University of Oxford (2000) 213–237
102

9. Dasgupta, P.: Trust as a Commodity. In: Trust: Making and Breaking Cooperative Relations.
Department of Sociology, University Oxford (2000) 49–72
10. Beth, T., Borcherding, M., Klein, B.: Valuation of trust in open networks. In: Proc. 3rd
European Symposium on Research in Computer Security – ESORICS ’94. (1994) 3–18
11. Mui, L., Mohtashemi, M., Halberstadt, A.: A computational model of trust and reputation.
In: Proceedings of the 35th Annual Hawaii International Conference on System Sciences
(HICSS’02)-Volume 7. (2002)
12. Maurer, U.: Modelling a public-key infrastructure. In: ESORICS: European Symposium on
Research in Computer Security, LNCS, Springer-Verlag (1996)
13. Abdul-Rahman, A.: The pgp trust model. EDI-Forum: the Journal of Electronic Commerce
(1997)
14. Joseph, S.: Neurogrid: Semantically routing queries in peer-to-peer networks. In: Revised
Papers from the NETWORKING 2002 Workshops on Web Engineering and Peer-to-Peer
Computing, Springer-Verlag (2002) 202–214
15. Reagle, J.: Trust in a cryptographic economy and digital security deposits:protocols and
policies. Master’s thesis, University of Maryland Baltimore County (1996)
16. Jøsang, A.: The right type of trust for distributed systems. In: Proceedings of the 1996
workshop on New security paradigms, ACM Press (1996) 119–131
17. Marsh, S.: Formalising Trust as a Computational Concept. PhD thesis, Department of Math-
ematics and Computer Science, University of Stirling (1994)
18. Ellison, C., Frantz, B., Lampson, B., Rivest, R., Thomas, B., Ylonen, T.: Spki certificate
theory. RFC 2693 (1999)
19. Housley, R., Ford, W., Polk, W., Solo, D.: Internet x.509 public key infrastructure certificate
and crl profile. RFC 2459 (1999)
20. OpenPrivacy: (Sierra: An openprivacy reputation management framework) OpenPrivacy
projects.
21. Chen, R., Yeager, W.: Poblano: A distributed trust model for peer-to-peer networks. Sun
Microsystems (2001)
103