Ant　Colony　Optimization　(ACO)　algorithm　is　a　meta-heuristic　and　stochastic　search　technology,　which　has　been　one　of　the　effective　tools　for　solving　discrete　optimization　problems.　However　there　are　two　bottlenecks　for　large-scaled　optimization　problems:　the　ACO　algorithm　needs　too　much　time　to　convergent　and　the　solutions　may　not　be　really　optimal.　This　paper　proposes　a　novel　ACO　algorithm　for　the　Multidimensional　Knapsack　Problems　(MKP),　which　employs　a　new　pheromone　diffusion　model　and　a　mutation　scheme.　First,　in　light　of　the　preference　to　better　solutions,　the　association　distances　among　objects　are　mined　in　each　iteration　with　Top-k　strategy.　Then,　a　pheromone　diffusion　model　based　on　info　fountain　of　an　object　is　established,　which　strengthens　the　collaborations　among　ants.　Finally,　an　unique　mutation　scheme　is　applied　to　optimizing　the　evolution　results　of　each　step.　The　experimental　results　for　the　benchmark　testing　set　of　MKPs　show　that　the　proposed　algorithm　can　not　only　get　much　more　optimal　solutions　but　also　greatly　enhance　convergence　speed