In　order　to　detect　the　attacks　on　non-control-data　in　a　program,　this　study　presents　a　data　flow　behavior　constraint　model　for　conditional　branch　decision-making　variables　in　a　program.　Our　model　analyzes　the　relationship　among　data　flow　and　decision-making　variables　and　verifies　them　during　executing,　to　ensure　the　conditional　branch　decision　statements　in　a　program　can　follow　their　expected　behavior　and　detecting　the　attack　on　branch　decision　variable　easily.　We　focus　on　the　constraints　including　the　collection　of　decision-making　variables　and　statements　that　the　decision-making　data　depending　on.　Program　slicing　technology　is　adopted　to　obtain　variables　and　statements　collections　which　are　dependent　on　decision-making　variables.　Then,　dynamic　incrementing　method　is　used　to　verify　the　program＇s　execution　dynamically.　By　comparing　whether　the　actual　behavior　is　consistent　with　the　expected　behavior,　we　can　learn　whether　the　program　is　attacked.　Experimental　results　indicate　that　our　method　can　be　effective　in　detecting　the　attacks　on　conditional　branch　decision-making　variables.　©　2011　by　Binary　Information　Press.