Generating realistic workloads for network intrusion detection systems

While the use of network intrusion detection systems (nIDS) is becoming pervasive, evaluating nIDS performance has been found to be challenging. The goal of this study is to determine how to generate realistic workloads for nIDS performance evaluation. We develop a workload model that appears to provide reasonably accurate estimates compared to real workloads. The model attempts to emulate a traffic mix of different applications, reflecting characteristics of each application and the way these interact with the system. We have implemented this model as part of a traffic generator that can be extended and tuned to reflect the needs of different scenarios. We also present an approach to measuring the capacity of a nIDS that does not require the setup of a full network testbed.