Research Of Anti-forensics Based On Digital Image Tampering

As the use of digita l multimedia content s uch as ima ges and video has increased,the forgers have the means and the incentive to create digital forgeries. Presently,powerful editing software allows forgers to create perceptua lly convinc ing digita lforgeries. Accordingly, there is a great need for techniques capable of authe nticatingdigital multimedia content. In response to this, researchers have begun develop ingdigital forens ic techniques capable of identifying d igita l forgeries. These forens ictechniques operate by detecting imperceptib le traces le ft by editing operations indigital multimed ia content.However, we cons ider the prob lem of multimedia security from the forger’s pointof view. We demonstrate that a n inte llige nt forger can design anti-forens ic operationsto hide editing fingerprints and fool the forens ic techniq ues. The research work ofanti-forensics can revea l the weakness of the existing forens ic algorithms. It can alsoencourage the produce of forens ics with more security and more robustness. The n itwill further promote the deve lopment of multimed ia forens ics. We start ouranti-forensic research fro m the existing multimed ia passive forens ics. The ma in worksare summarized as follows:Firstly, an anti-forens ic sche me for USM Sharpening is proposed to inva lidatethe existing forens ic a lgorithms. It is s imp le but effective. The contributio n of theproposed approach lies in the introduction of d ithering no ise to the overshoot ima ges.It remo ves the overshoot artifacts in image edges and abrupt change in histogra m ends.The effective ness of the proposed method is proved by the experime nta l results on alarge image database with various parameter settings. Comparisons are made amongthe unsharpened images, the sharpened images and the a nti-forens ic dithered ima ge.Both the detection ability and ima ge quality are used for its performa nce eva luation.Secondly, we proposed an anti-forens ics method to inva lidate the noise-resid uebased passive forens ics techniq ues. With the operation of deleting moving objects, forexamp le, video tampering changes the statistica l distribution of no ise resid ue. Thenoise residues are extracted fro m the inp ut video and intentio nally added into thetampered video, so as to guarantee that vide o tampering will not cha nge the te mporalcorrelation o f noise res idue. The c lue for passive forens ics is hided, and thusanti-forensics is achieved. The experimenta l results on object-based video tampering show that the proposed anti-forens ics approach is effective to fool those noiseresidue-based passive forens ics techniq ues.At present, the multimed ia anti-forens ics research is still in its preliminaryexploration phase. The rele vant literatures are quite few. However, it is a subject withgreat research va lue, because it can promote the further develop ment o f pass iveforens ics techno logy. Aiming at the two passive forens ics a lgorithms in image/videoforens ic research, preliminary results are obta ined.