MADneSs: A Multi-Layer Anomaly Detection Framework for Complex Dynamic Systems

Anomaly detection can infer the presence of errors without observing target services, but detecting variations in observable parts system on which services reside. This is a promising technique complex software-intensive systems, because either instrumenting services' internals exceedingly time-consuming, or encapsulation makes them not accessible. Unfortunately, such systems anomaly often ineffective due to their dynamicity, implies changes expected workload. Here we present our approach enhance efficacy complex, dynamic systems. After discussing related challenges, MADneSs, an framework tailored for above that includes adaptive multi-layer monitoring module. Monitored data are then processed by detector, adapts its parameters depending current behavior. An alert provided if analysis conducted detector identify unexpected trends data. MADneSs evaluated through experimental campaign two service-oriented architectures; software faults injected application layer, and detected underlying layers. Lastly, quantitatively qualitatively discuss results with respect state-of-the-art solutions, highlighting key contributions MADneSs.