This　paper　analyzes　the　optimal　strategies　for　the　attacker　and　the　defender　in　an　attack-defense　game,　considering　the　risk　attitudes　of　both　parties.　The　defender　moves　first,　allocating　its　limited　resources　to　three　different　measures:　launching　a　proactive　strike　or　preventive　strike,　building　false　targets,　and　protecting　its　genuine　object.　It　is　assumed　that　(a)　launching　a　proactive　strike　has　limited　effectiveness　on　its　rival　and　does　not　expose　the　genuine　object　itself,　(b)　a　false　target　might　be　correctly　identified　as　false,　and　(c)　launching　a　preventive　strike　consumes　less　resources　than　a　proactive　strike　and　might　expose　the　genuine　object.　The　attacker　moves　after　observing　the　defender＇s　movements,　allocating　its　limited　resources　to　three　measures:　protecting　its　own　base　from　a　proactive　strike　or　preventive　strike,　building　false　bases,　and　attacking　the　defender＇s　genuine　object.　For　each　of　the　defender＇s　given　strategies,　the　attacker　chooses　the　attack　strategy　that　maximizes　its　cumulative　prospect　value,　which　accounts　for　the　players＇　risk　attitudes.　Similarly,　the　defender　maximizes　its　cumulative　prospect　value　by　anticipating　that　the　attacker　will　always　choose　the　strategy　combination　that　maximizes　its　own　cumulative　prospect　value.　Backward　induction　is　used　to　obtain　the　optimal　defense,　attack　strategies,　and　their　corresponding　cumulative　prospect　values.　Our　results　show　that　the　introduction　of　risk　attitudes　leads　the　game　to　a　lose-lose　situation　under　some　circumstances　and　benefits　one　party　in　other　cases.