Ries, Sebastian (2009)
Trust in Ubiquitous Computing.
Technische Universität Darmstadt
Dissertation, Erstveröffentlichung
Kurzbeschreibung (Abstract)
In the vision of ubiquitous computing, the activities of daily life are supported by a multitude of heterogeneous, loosely coupled computing devices. The support of seamless collaboration between users, as well as between their devices, can be seen as one of the key challenges for this vision to come true. This thesis provides a trust based approach to supporting the selection of trustworthy interaction partners. The goal of this approach is to estimate an entity's trustworthiness as accurately as possible in order to improve the average quality of the entity's interactions. In this thesis, the trustworthiness of an entity is derived from evidence gained during past interactions. To this end, current Bayesian trust models are extended and improved regarding the following aspects: (i) better integration of the characteristics of the application context, (ii) more intuitive access to the trust model, and (iii) better integration of recommendations by third parties. The last aspect is important as there are numerous situations in which direct evidence between entities is rare. The proposed approach provides means for the robust integration of recommendations provided by third parties, especially considering attacks by entities intentionally providing misleading recommendations. Scientific Contribution: The scientific contribution of this thesis is summarized as follows: The trust model that is provided in this thesis extends Bayesian trust models in order to improve the integration of context-dependent parameters, such as dispositional trust and aging of evidence. Furthermore, a parameter called maximum number of evidence units allows the user to define the number of evidence that is expected to be sufficient for being representative for an entity's behavior within a certain application context. In the proposed model, the dispositional trust can be assessed according to the preference of the user; alternatively, a new approach for deriving the dispositional trust from the behavior of previously encountered entities is provided. The proposed interrelation between the aging and the maximum number of expected evidence units allows the limitations of current Bayesian trust models to be overcome. The thesis shows that in those models, aging either does not have an impact on the expectation value in the absence of evidence, or it narrows the range of the expectation value. A second representation of trust - called the Human Trust Interface (HTI) - is proposed providing for an easier access to the model by human users. This representation is based on a simple set of parameters. These parameters are also the basis for a graphical representation allowing users to interpret and adjust the trust values of other entities intuitively. As the model supports two different representations a mapping between both representations is required in order to switch between both representations. The provided mapping allows users and developers of trust models to benefit from the advantages of both representations. The distributed computational model that is proposed for the aggregation of direct evidence and recommendations has been designed to be especially robust to so-called Sybil attacks, which occur when a single party tries to multiply the influence of its recommendations by creating a high number of seemingly independent entities. This is achieved using the accuracy of a recommender's past recommendations as well as the rank of the recommender in order to limit a recommender's influence. Especially, considering the rank of a recommender, i.e., its position in the group of recommenders, provides a means for limiting the influence of a potentially infinite number of malicious recommenders under certain circumstances. Evaluation: The trust model has been evaluated in two user studies which support that users feel comfortable with the proposed graphical representation. Furthermore, in the simulation of collaboration in an opportunistic network, the model shows a good performance regarding the estimation of an entity's trustworthiness and regarding the average quality of interactions when using the trust model to find the best interaction partner. This results from the comparison to a state-of-the-art approach, as well as from a comparison to an artificial model that is initialized with the system variables of the simulation model, and therefore serves as perfect selection strategy. The simulation shows the results of the different approaches over a set of 15 populations, which have been canonically derived from the system model, modeling entities with different typical behaviors.
Typ des Eintrags: | Dissertation | ||||
---|---|---|---|---|---|
Erschienen: | 2009 | ||||
Autor(en): | Ries, Sebastian | ||||
Art des Eintrags: | Erstveröffentlichung | ||||
Titel: | Trust in Ubiquitous Computing | ||||
Sprache: | Englisch | ||||
Referenten: | Mühlhäuser, Prof. Dr. Max ; Josang, Prof. Dr. Audun | ||||
Publikationsjahr: | 30 Oktober 2009 | ||||
Ort: | Darmstadt | ||||
Verlag: | Technische Universität | ||||
Datum der mündlichen Prüfung: | 2 Juli 2009 | ||||
URL / URN: | urn:nbn:de:tuda-tuprints-19486 | ||||
Kurzbeschreibung (Abstract): | In the vision of ubiquitous computing, the activities of daily life are supported by a multitude of heterogeneous, loosely coupled computing devices. The support of seamless collaboration between users, as well as between their devices, can be seen as one of the key challenges for this vision to come true. This thesis provides a trust based approach to supporting the selection of trustworthy interaction partners. The goal of this approach is to estimate an entity's trustworthiness as accurately as possible in order to improve the average quality of the entity's interactions. In this thesis, the trustworthiness of an entity is derived from evidence gained during past interactions. To this end, current Bayesian trust models are extended and improved regarding the following aspects: (i) better integration of the characteristics of the application context, (ii) more intuitive access to the trust model, and (iii) better integration of recommendations by third parties. The last aspect is important as there are numerous situations in which direct evidence between entities is rare. The proposed approach provides means for the robust integration of recommendations provided by third parties, especially considering attacks by entities intentionally providing misleading recommendations. Scientific Contribution: The scientific contribution of this thesis is summarized as follows: The trust model that is provided in this thesis extends Bayesian trust models in order to improve the integration of context-dependent parameters, such as dispositional trust and aging of evidence. Furthermore, a parameter called maximum number of evidence units allows the user to define the number of evidence that is expected to be sufficient for being representative for an entity's behavior within a certain application context. In the proposed model, the dispositional trust can be assessed according to the preference of the user; alternatively, a new approach for deriving the dispositional trust from the behavior of previously encountered entities is provided. The proposed interrelation between the aging and the maximum number of expected evidence units allows the limitations of current Bayesian trust models to be overcome. The thesis shows that in those models, aging either does not have an impact on the expectation value in the absence of evidence, or it narrows the range of the expectation value. A second representation of trust - called the Human Trust Interface (HTI) - is proposed providing for an easier access to the model by human users. This representation is based on a simple set of parameters. These parameters are also the basis for a graphical representation allowing users to interpret and adjust the trust values of other entities intuitively. As the model supports two different representations a mapping between both representations is required in order to switch between both representations. The provided mapping allows users and developers of trust models to benefit from the advantages of both representations. The distributed computational model that is proposed for the aggregation of direct evidence and recommendations has been designed to be especially robust to so-called Sybil attacks, which occur when a single party tries to multiply the influence of its recommendations by creating a high number of seemingly independent entities. This is achieved using the accuracy of a recommender's past recommendations as well as the rank of the recommender in order to limit a recommender's influence. Especially, considering the rank of a recommender, i.e., its position in the group of recommenders, provides a means for limiting the influence of a potentially infinite number of malicious recommenders under certain circumstances. Evaluation: The trust model has been evaluated in two user studies which support that users feel comfortable with the proposed graphical representation. Furthermore, in the simulation of collaboration in an opportunistic network, the model shows a good performance regarding the estimation of an entity's trustworthiness and regarding the average quality of interactions when using the trust model to find the best interaction partner. This results from the comparison to a state-of-the-art approach, as well as from a comparison to an artificial model that is initialized with the system variables of the simulation model, and therefore serves as perfect selection strategy. The simulation shows the results of the different approaches over a set of 15 populations, which have been canonically derived from the system model, modeling entities with different typical behaviors. |
||||
Alternatives oder übersetztes Abstract: |
|
||||
Freie Schlagworte: | Trust, Reputation, Trust modeling, Ubiquitous Computing | ||||
Sachgruppe der Dewey Dezimalklassifikatin (DDC): | 000 Allgemeines, Informatik, Informationswissenschaft > 004 Informatik | ||||
Fachbereich(e)/-gebiet(e): | 20 Fachbereich Informatik > Telekooperation 20 Fachbereich Informatik |
||||
Hinterlegungsdatum: | 30 Okt 2009 13:27 | ||||
Letzte Änderung: | 05 Mär 2013 09:28 | ||||
PPN: | |||||
Referenten: | Mühlhäuser, Prof. Dr. Max ; Josang, Prof. Dr. Audun | ||||
Datum der mündlichen Prüfung / Verteidigung / mdl. Prüfung: | 2 Juli 2009 | ||||
Export: | |||||
Suche nach Titel in: | TUfind oder in Google |
Frage zum Eintrag |
Optionen (nur für Redakteure)
Redaktionelle Details anzeigen |