Abstract:  A strategy selection model for network survivability based on three players’ dynamic game is proposed to efficiently improve the active defensive ability of a network in the face of attack, defender, and failure accidents. The non-cooperative complete information dynamic game theory is used to construct the
three players’ survivability game model which takes attacker, defender and accidents as the game’s participators.The formal expression of sub-game perfectness Nash equilibrium by backward induction is given.The strategy selection algorithm for three players’ dynamic game is then proposed. Simulations show that each optimal strategy acquired by the given strategy selection model and the algorithm can meet the actual requirement.

Key words: network survivability, dynamic game, strategy selection model, backward induction