Monitoring Power Usage Effectiveness to Detect Cooling Systems Attacks and Failures in Cloud Data Centers

Energy-related Denial of Service (DoS) attacks have the potential to impact not only the quality or availability of services provided by large-scale data center (DC) infrastructures but also the operating expenses incurred by the organizations managing them, essentially in terms of energy bills. More specifically, the impact on the overall energy usage and, consequently, on the associated expenses, increases with the amount of time required to identify the attack. Therefore, the degradation of the environmental control systems in the buildings/facilities hosting the computing or storage nodes poses an especially insidious threat, which could result in a novel kind of attack to the involved infrastructures. Due to the limited ability to observe events in cyber-physical systems, recognizing these violations is extremely challenging for data center administrators. This paper proposed a new methodology for detecting cooling systems attacks based on continuous Power Usage Effectiveness (PUE) monitoring. This kind of measurement is quite simple to arrange in a data center, and can help to detect, in a limited amount of time, both attacks and failures on a DC cooling system, thus helping the system administration to limit expenses and service outages.