Networks are present in virtually all aspects of life. The world surrounding usincludes to many networks. For example, communication networks constituted of phones,electrical networks, computers networks, aerial lines network, ? etc, are such importantnetworks in our daily life. The mathematical framework of networks is well appropriatedto describe different systems composed of many entities interacting with each other. Eachentity is represented by a network node and each interaction by a link between twonodes. Therefore, it is possible to model these networks by graphs. For most of thesenetworks, the difficulty comes mainly from the large number of entities and the way theyare interconnected. A natural approach to simplify such systems is therefore to reducetheir size. This simplification is not made randomly, but in such a way that the nodes ofthe same component would have more connections between themselves than with othercomponents. These groups of nodes or components are called communities of interest.Our thesis is positioned in the field of social graphs study. It is mainly interested instudying the robustness of social structures emerging in interaction networks. The aspectof networks robustness is a very important challenge to understand their functioning,the behavior of the constituting entities and especially to understand the interactionsthat may occur between them, allowing the emergence of certain behaviors that were notpredictable at all in advance. Currently, studies of networks robustness that exist in theliterature treat this aspect from a purely structural point of view, ie, all perturbations areapplied either on nodes or on the edges of the graph. In terms of our study, we focused ondefining a new strategy based on perturbations applied on the parameters that allow theemergence of social graphs in interaction networks. This way to approach the robustnessappearance of the graphs is a new way to assess and quantify the changes that may occurin the structures of these graphs.
