چكيده به لاتين
Removing nodes or links from a real-world social network may lead to a collapse in the entire network itself. This is because of the propagation effect of the initial removal. In the literature, this phenomenon is called cascading failure. In the context of trust modeling, a cascading failure phenomenon occurs when a node’s trust toward another one changes to distrust. This change in the trust network may impact the other nodes’ trust toward the target node and changes their trust opinion as well. As the number of failures in a network increases and the trust in the entire network decreases, the users become more reluctant to share their interests and opinions with other members. Currently, simple computational trust models are used in the literature for modeling cascading trust failures. The effect of relevant contexts in modeling trust and cascading trust failures is missed in the proposed models so far. Failure in a specific trust context may impact the relevant and dependent contexts as well. It seems to be necessary to have a more complete and comprehensive trust modeling approach besides modeling just the cascading trust failures. In this thesis, the computational trust is formulated by considering the context dependencies in addition to the impact of trust contexts on one another. Also, by mapping trust contexts to multiplex networks' layers and using the advantages of complex network analysis concepts, a new method for computing the similarity between the trust contexts is introduced. The introduced trust model uses the trust information of all layers (i.e. contexts) to compute trust values of the next step. In addition, the trust model uses the newly provided information to adjust the computed trust value with the help of real-world data. The trust cascading failure model and the attack prevention method are introduced as well. Through a series of evaluation scenarios, the proposed model's ability in detecting missing trust links is evaluated. It is shown that the model is able to predict almost 70 percent of missing links when the contexts are similar. The real-world networks' data, such as Facebook's Egonets and simulated data are used for the analysis. The multiplex networks are made up of multiple single networks. It is shown that the higher values for the context importance parameter make the trust links more vulnerable and easier to fail. The three well-known trust attack scenarios including HT, LT and RT are performed and it is demonstrated that the layers with high similarity values tend to have more similar cascading failures process. By adding the attack prevention component, the model's accuracy gets close to 0.9, which is a notable improvement.
