Une nouvelle approche pour la détection d’anomalies dans les flux de graphes hétérogènes

In this work, we propose a new approach to detect anomalous graphs in a stream of di- rected and labeled heterogeneous graphs. Our approach uses a new representation of graphs by vectors. This representation is flexible and allows to update the graph vectors as soon as a new edge arrives. In addition, it is applicable to any type of graph and optimizes memory space. Moreover, it allows the detection of anomalies in real-time.