This　paper　analyses　the　optimal　strategies　for　one　defender　and　two　attackers　in　a　defence-attack　game,　where　a)　the　defender　allocates　its　resource　into　defending　against　and　attacking　the　two　attackers,　and　b)　the　two　attackers,　after　observing　the　action　of　the　defender,　allocate　their　resources　into　attacking　and　defending　against　the　defender,　on　either　a　cooperative　or　non-cooperative　basis.　On　a　cooperative　basis,　for　each　of　the　defender＇s　given　strategies,　the　two　attackers　work　together　to　maximise　the　sum　of　their　cumulative　prospect　values　while　anticipating　the　eight　possible　game　outcomes.　On　a　non-cooperative　basis,　for　each　of　the　defender＇s　given　strategies,　each　attacker　simultaneously　yet　independently　tries　to　maximise　their　own　cumulative　prospect　value.　In　both　cases,　the　defender　maximises　its　cumulative　prospect　value　while　anticipating　the　attackers＇　actions.　Backward　induction　is　employed　to　obtain　the　optimal　defence　and　attack　strategies　for　all　scenarios.　Numerical　examples　are　performed　to　illustrate　the　applications　of　the　strategies.　In　general,　we　find　two　opposing　effects　considering　the　attackers＇　strategies　and　analyse　the　alteration　of　strategies　for　the　participants　under　two　different　risk　preferences:　risk-averse　and　risk　seeking.　The　reasons　for　the　alteration　are　also　performed　to　illustrate　the　practical　applications.