Railway　network　is　one　of　the　busiest　regional　transportation　infrastructures,　which　is　exposed　to　a　high　risk　of　intentional　attacks.　Given　the　railway　network　stations　have　a　larger　service　area,　attackers　may　have　different　biases　toward　the　valuation　of　railway　stations　or　lines.　This　paper　proposes　a　method　for　optimally　allocating　defensive　resources　based　on　a　Bayesian　game　model　and　a　comprehensive　importance　evaluation　model　of　stations　by　multi-layer　network　models,　aiming　to　reduce　the　losses　of　defenders.　The　attack　strategy　was　made　according　to　the　importance　of　railway　stations　evaluated　by　three-layer　network　models,　namely　topology　layer,　the　ridership　layer　and　the　travel　time　layer,　which　depict　the　features　of　railway　networks　and　also　reflect　the　variety　of　attacker＇s　biases.　The　optimal　allocation　of　defensive　resources　was　obtained　under　the　Nash　equilibrium　of　Bayesian　game.　The　proposed　method　is　implemented　in　a　regional　railway　network　in　north　China,　and　the　case　network＇s　risk　under　various　attack　strategies　were　compared　to　validate　the　applicability　of　this　model.　The　application　results　show　that　the　optimal　defensive　resources　allocation　based　on　the　importance　evaluation　by　three-layer　models　has　the　lowest　risk　considering　the　variety　in　the　attacker＇s　biases.