Network Embedding Attack: An Euclidean Distance Based Method

Network embedding methods are widely used in graph data mining. This chapter proposes a Genetic Algorithm (GA) based Euclidean Distance Attack strategy (EDA) to attack the DeepWalk-based network embedding to prevent certain structural information from being discovered. EDA disrupts the Euclidean distance between pairs of nodes in the embedding space by making a minimal modification of the network structure, thereby rendering downstream network algorithms ineffective, because a large number of network embedding based downstream algorithms, such as community detection and node classification, evaluate the similarity based on the Euclidean distance between nodes. Different from traditional attack strategies, EDA is an unsupervised network embedding attack method, which does not need labeling information.