Caracterização e Identificação de Papéis e Vértices em Redes de Conexão e Redes Aleatórias

The structure of different networks, such as social networks is important beause their study can lead to the understanding of various phenomena. The public availability of these networks, however, has enhanced the difficulty of preserving the users’ privacy whose PII (Personally Identifiable Information) are likely tobe revealed given the weaknesses of current anonymization techniques. This work resents two aspects are done related to the problem of re-identification of verticesin anonymized networks. The first one is re-identification of roles in connection networks of anonymized TCP traffic. In this network, the possible vertex roles are Client, Server and Client/Server depending on the connection open patterns. We present a characterization of roles and network structure and based on these obserations, we propose three techniques for roles re-identification using only structural properties of anonymous connection networks. Numerical evaluation shows that it is possible to re-identify the role of a high fraction of vertices (more than 95%, in somecases). The second work is an investigation of the structural distinction of vertices through unique values of network properties. We consider random graph models such as the G(n, p) and preferential attachment and different structural propertie such as degree and betweenness. The results indicate that in all models studied there are vertices that are uniquely identifiable by some structural property forsome region of parameter space.