F5隐写算法及其隐写分析研究

8840[1]2000A.WestfeldF5[2]F52002FridrichF5[3]1F51.1R.Crandall[2]LSB50%LSB25%3.12%LSBRS[4]217F41157F54593.8nLSBk(nk)3LSBa1,a2,a3x1x2LSB4x1=a1a3x2=a2a3x1a1a3x2=a2a3a1x1=a1a3x2a2a3a2x1a1a3x2a2a3a341max(max,,)F5max=1(1,,)=21LSBLSB=1+1=12(1)==1+1=21(2)[e-mail](03F51060)(418010703)(1963)(1955)(1965-)F51,212(1.3430092.100083)WestfeldF5F5F5F5;;;;F5;TP393.08A1000-7024(2007)04-0884-04StudiesofF5steganographicalgorithmanditssteganalysisLIUChang-xin1,2,ZHUPing1,XIAChun-he2(1.DepartmentofComputerScience,JinggangshanCollege,Ji'an343009,China;2.SchoolofComputerScience,BeihangUniversity,Beijing100083,China)AbstractImagesteganalysisisatechniqueofdetecting,extractinganddestroyingthehidinginformationinstegoimage,andithasbecomeoneofthefocusesinthefieldofinformationsecurity.F5SteganographicAlgorithmisaimportantsteganography.F5Stegano-graphicAlgorithmanditsSteganographicsystemarepresented.AsteganalyticmethodthatdetectmessageshiddeninJPEGimagesisdiscussed.ExperimentalresultsshowthatF5algorithmimplementsmatrixencodingtoimprovetheefficiencyofembedding.Keywordssteganography;coverimage;stegoimage;matrixencoding;F5algorithm;steganalysis20072Feb.2007284Vol.28No.4ComputerEngineeringandDesign885==221(3)1100050000=1000/50000=2%1(=8)(=9)=850000:255=196=255(1,255,8)196*8=1568=91.2F5=(1,2,,)F5F5(1)QDCTJPEG(2)PRNG(3)PRNGDCT(4)(5)=21(6)(1,,)ACHash==1Hash=,1Hash=1=,=0=1,2,,¬,,001(7)JPEGHuffman(8)1.3F5F5[2]12F5JPEG13%F5LSB200*200lena.jpg3237223312341kn(%)(%)12345678913715316312725551150.0025.0012.506.253.121.560.780.390.20100.0066.6742.8626.6716.139.525.513.141.7622.673.434.275.166.097.068.039.021F5e,p,q,cOption=eOption=pOption=q80Option=cemdFileNamePasswod=Quality=Interger.parseInt(args[i+1])Comment=File=inFileNameimage=bmp.getImage()image=Toolkit.getDefaultToolkit().getImage(inFileName)Jpegjpg.Compress()JpegJpegEncoder(image,Quality,dataOut,comment)StandardUsageYNNdeZigZag2F5carrier=newbyte[(int)f.length()]fis=newFileInputStream(f)fos=newFileOutputStream(newFile(embFileName))fis.read(carrier)Huffmanhd=newHuffmanDecode(carrier)coeff=hd.decode()F5Random(password.getBytes())random,Permutation(coeff.length,random)01fos.write(byte)extractedByte8861.58772F52002J.FridrichF5[3,5,6]2.1F52.1.1F534BB3*3B22=1-4eB21=B23=B12=B32=e05()52.1.2F5DCTDCT=/=(1)+(2)+(3)=1++100=0+1=0(4)(4)=argmin[001]2+[(1)112]2(5)(5)=100+112121+212(6)DCT(,){(1,2),(2,1),(2,2)}=(12+21+22)/3(7)2.1.3(1,2k-1,k)W(k)M==2211=22111/=2211(8)=00,=1+28(9)2.2F5F5(9)AC(7)=*=21(1)ACDCT1DCT(1)(8)2.3J.FridrichQF7520JPEG2345F540003500300025002000150010005000FrequencyofOccurrence-8-6-4-202468estimatedhistogramValuesoftheDCTCoefficient(2,1)coverimagehistogramstegoimagehistogram2DCTimage''123456789101112131415161718192000.20200.2590.2440.2340.2160.3470000.3420.499000.257000.3700.4280.1060.2380.0790.2730.2560.2760.2480.4090.0440.0700.1030.2500.5220.1130.0780.2910.0830.0730.3290.377198452025421401202671967524741235892277520164239302427811846219375214194902101122040211762587325705124618715745215318386457120955722245131934219308887[0.20.5]0.13A.WestfeldF5J.FridrichF5F5LSBRSJ.FridrichJ.Fridrich[7~9]:[1]WestfeldA,PfitzmanA.Attacksonstegographicsystem[C].Pro-ceedingsof3thInternationalWorkshoponInformationHiding,2001.517-522.[2]WestfeldA.F5-Asteganographicalgorithm:Highcapacitydes-pitebettersteganalysis[C].NewYork,Berlin,Heidellberg:Springer-Verlag,2001.289-302.[3]FridrichJ,GoljanM,HogeaD.SteganalysisofJPEGimages:BreakingtheF5algorithm[C].Netherlands:5thInformationHi-dingWorkshop,2002.7-9.[4]FridrichJ,GoljanM,DuR.DetectingLSBsteganographyincolorandgray-scaleimages[J].MagazineofIEEEMultimedia:SignalIssueonSecurity,2001,8(4):22-28.[5]FridrichJ,GoljanM,HogeaD.NewmethodologyforbreakingsteganographictechniquesforJPEGs[C].CA:ProcEISPIESantaClara,2003.143-155.[6]FridrichJ,SoukalD.Quantitativesteganalysisofdigitalimages:Estimatingthesecretmessagelength[J].ACMMultimediaSys-temsJournal,2003,9(3):288-302.[7]FridrichJ,GoljanM.OnestimationofsecretmessagelengthinLSBsteganographyinspatialdomain[C].CA:ProcEISPIESanJose,2004.[8]FridrichJ,SoukalD,GoljanM.MaximumlikelihoodestimationofsecretmessagelengthembeddedusingPMKsteganographyinspatialdomain[C].CA:ProcEISPIESanJose,2005.595-606.[9]FridrichJ,HolotyakTS,SoukalD.Stochasticapproachtosecretmessagelengthestimationinkembeddingsteganography[C].CA:ProcEISPIESanJose,2005.673-684.(883)4:[1]KimD-S,LeeW-H,KweonI-S.Automaticedgedetectionusing3×3idealbinarypixelpatternsandfuzzy-basededgethreshol-ding[J].PatternRecognitionLett,2004,25:101-106.[2]RosinPL,IoannidisE.Evaluationofglobalimagethresholdingforchangedetection[J].PatternRecognitionLett,2003,24:2345-2356.[3]FengL,SuenCY,TangYY,etal.Edgeextractionofimagesbyre-constructionusingwaveletdecompositiondetailsatdifferentresolutionlevels[J].PatternRecognitionandArtificalIntelli-gence,2000,14(6):779-793.[4]BrainMSadler,SwamiA.Analysisofmultiscaleproductsforstepdetectionandestimation[J].IEEETransInformationTheory,1999,45:1043-1051.[5]ZhangQ.Waveletnetworks[J].IEEETransonNeuralNetworks,1992,3(6):889-898.[6]ChoiH