Το περιβάλλον-πλαίσιο που αναλύεται και αναπτύσσεται στην παρούσα διπλωματική επιτρέπει σε χρήστες να λαμβάνουν μια εκτίμηση της εμπιστοσύνης, που λαμβάνει μια υπηρεσία, ή ένας πόρος, που παρέχεται από συγκεκριμένο πάροχο για συγκεκριμένο περιεχόμενο. Συγκεκριμένα, ο χρήστης ορίζει τις απαιτήσεις μου έχει από την υπηρεσία και στην συνέχεια, αφού υπολογιστεί η πιθανοτική τιμή ικανοποίησης της κάθε απαίτησης, υπολογίζεται, λαμβάνοντας υπόψιν και τις αλληλεπιδράσεις, η τελική τιμή της εμπιστοσύνης.
(EL)
The main goal of this diploma thesis is the development of a framework that allows users to
receive an
assessment of the trust a service, provided by a single provider o
n specific context,
is
entitled to. Given the increased frequency online services are used, the need for swift and
efficient evaluation of their reliability increases respectively. The fram
ework is designed to cover
this need by assessing the extend to which a service in interest is trusted, keeping in mind the
user’s preferences, regarding requi
rements from the service on one hand, and external
information, associated with each individual r
equirement, and its level of satisfaction, on the
other.
For each and every requirement, that is used in the framework, a definition is required. The
definition consists of external information, that have an impact on the requirement in question, as
well
as their respective correlations. The framework models the user’s needs and, subsequently,
creates a goal tree, representing each requirement. Using logic inference techniques, relying
heavily on Markov Logic Networks,
an appreciation of a requirement
’
s sa
tisfiability is
calculated. Calculation can be performed with lack of information as well, which is an advantage
of this method.
A definition of the overall trust, and the requirements from which it consists of, is, moreover,
required by the framework. A g
raph corresponding to the given definition is constructed, based
on certain criteria. The TidalTrust algorithm, which was initially proposed for social networks, is
then applied upon this graph. The algorithm’s result value is provided to the user as an
as
sessment of the trust
worthiness of the service, which is evaluated.
Alterations on information and models can emerge later. The framework responds to these
changes by discovering them, at first, and reinferring trust on a service, wherever it’s deemed
nece
ssary, afterwards. Similar behavior is portrayed when the time frame of information validity
expires. This frame is supplied to the framework by the user
,
and is responsible for renewing the
provided information
,
in specific time intervals.
The framework’s
architecture, which is defined and materialized, as part of this diploma thesis,
is based on the Blackboard architectural pattern, as well as the Publisher/Subscriber architectural
pattern. Such a choice enables the decoupling among the discrete stages of
the trust inference
process, and thus the possibility of executing the stages that are absolutely essential for the
service assessment, given the alterations that have occurred at that certain point
(EN)
National Technical University of Athens
