Detection and Discrimination of Injected Network Faults

Although the present work does in fact employ training data, it does so in the interest of calibrating the results Six hundred faults were induced by injection into five live obtained from an experimental detection and diagnostic campus networks at Carnegie Mellon University in order system designed specifically to accommodate noisy, to determine whether or not particular network faults nonstationary, nonspecific domains. The system have unique signatures as determined by out-of-band generalizes by virtue of its log analysis capabilities; all monitoring instrumentation. If unique signatures span monitored data and events are recorded in log files. networks, then the monitoring instrumentation can be These files are processed by the system, resulting in used to diagnose network faults, or distinguish among testable and reproducible detections and diagnoses of fault classes, without human intervention, using anomalous conditions. Any monitored process or machine-generated diagnostic deci...

• Type: Conference
• Year: 1993
• Where: FTCS
• Authors: Roy A. Maxion, Robert T. Olszewski
• Keywords: FTCS 1993, Network Faults, Particular Network Faults, Unique Signatures