In this paper, we present preliminary results of analyzing data from the Dark Web collection using a dynamical systems approach for unsupervised anomaly detection. The goal is to provide a robust, focus-of-attention mechanism to identify emerging threats in time-dependent, unlabelled data sets. In our method, finite-time Lyapunov exponents are used to characterize the time evolution of both the directed network structure and the distribution of text attributes in the forum messages. We provide a description of the technique and a summary of the initial anomaly detection results. We conclude with a summary of results and a brief discussion of promising avenues for future research.

No abstract

No abstract

No abstract