This　paper　proposed　a　JPEG　steganalysis　scheme　based　on　Markov　model　using　features　derived　from　the　detected　image　and　the　estimated　image.　Estimated　image　is　created　by　cropping　four　pixels　of　the　detected　image　from　left　line.　And　the　estimated　image　is　similar　with　the　original　image　with　the　statistical　characteristics.　From　both　of　the　detected　image　and　estimated　image,　Markov　process　is　applied　to　modeling　the　difference　JPEG　2-D　arrays　along　horizontal,　vertical,　and　diagonal　directions　so　as　to　utilize　the　high　order　statistics　for　enhancing　changes　caused　by　JPEG　steganography.　Support　vector　machines　(SVM)　are　utilized　as　classifier.　The　experimental　results　have　proved　that　the　proposed　method　is　effective　in　attacking　by　the　existing　steganalyzers.