Abstract:

A mathematical model of convergence time for probabilistic packet marking is established. In the model, the attack path reconstruction process is described as a convergent stochastic process. Probability distribution of convergence time and the quantitative relationship among convergence time, path length and marking probability are inferred. The minimum condition of convergence time is obtained. On the basis of this model, two improved mathematical models of convergence time for two traditional probabilistic packet marking approaches are presented. The simulation results show that the improved models are more accurate than the given two traditional models.

Key words: network security;denial of service;IP traceback;probabilistic packet marking