 |
Center for Nonlinear Studies |
In the first part of the talk we consider a game in which a defender of network classifies an intruder as a spy or a spammer. The classification is based on a finite number of observations of attacks of two different targets. The spammer is non-strategic and attacks randomly (with a known distribution). The spy strategically selects the number of attacks on his main target: the file server. The defender strategically selects his classification policy: a threshold on the number of file server attacks. The defender needs to balance missed detections and false alarms, while the spy has a tradeoff between attacking the file server more aggressively and increasing the chance of getting caught.
In the second part of the talk we study a game theoretic model of competing network service providers that are connected in parallel and serial combinations and that strategically price their service in the presence of elastic user demand. To obtain our results, we make an analogy between the game and an electric circuit, in which the slope of each provider’s latency function is made to be analogous to electrical resistance. Our bound on efficiency loss depends on the ratio of the conductance of the circuit branch with highest conductance to the combined conductance of all the branches. In terms of the original problem, the bound measures how the worst-case efficiency loss increases as the capability of the system becomes more concentrated in a particular serial combination of players.
Host: David Wolpert, CCS-3, 665-7914, Game Theory Seminar Series