Complex systems experience dramatic changes in behavior and can undergo transitions from functional to dysfunctional states. An unstable system is prone to dysfunctional collective cascades that result from self-reinforcing behaviors within the system. Because many human and technological civilian and military systems today are complex systems, understanding their susceptibility to collective failure is a critical problem. Understanding vulnerability in complex systems requires an approach that characterizes the coupled behaviors at multiple scales of cascading failures. We used neuromorphic methods, which are modeled on the pattern-recognition circuitry of the brain and can find patterns in high-dimensional data at multiple scales, to develop a procedure for identifying the vulnerabilities of complex systems. This procedure was tested on microdynamic Internet2 network data. The result was a generic pipeline for identifying extreme events in high dimensional datasets.