The　core　of　future　network　security　protection　will　be　active　defence,　where　security　situation　prediction　plays　an　important　part.　In　order　to　solve　the　problem　of　uncertainty　conversion　between　qualitative　concept　and　quantitative　values　in　situational　awareness,　and　to　improve　the　degree　of　matching　between　data　distribution　and　actual　situation,　this　paper　presents　a　posture　prediction　algorithm　based　on　multidimensional　cloud　model.　Several　cloud　models　are　constructed　according　to　training　historical　data,　and　obtain　the　corresponding　integrated　cloud　of　key　attributes　by　soft-or　calculation.　At　last　we　get　network　future　situation　value　in　the　inverse　operations　of　membership　function.　Experiments　show　that　the　method　has　good　performance　of　detection　accuracy　predict　trends　stability.　©　2012　IEEE.