Cost-sensitive security risk management for large-scale computing infrastructures

Security threats to large-scale computing and networking infrastructures are increasingly prevalent and difficult to manage. Because of cost, labor, and other operational constraints present in large computing systems, it is often not possible to address all security risks immediately as they are discovered. This leaves known security threats unaddressed and risk managers are forced to dynamically prioritize different classes of threats. We use techniques from optimal control to derive a dynamic policy for choosing different security modes in a manner that achieves security management goals in a cost-sensitive manner. By leveraging results from the applied probability literature, we show that under some very general conditions our proposed scheme will keep unaddressed security risks at acceptable levels whenever possible.