The Research On Steganalysis Method For Vector Graphics Based On The Statistics Of Geometric Features

With the development of the multimedia and information technology,information security is becoming an increasingly prominent issue. Currently,Information hiding is a research hotspot in the digital media security. It can embedinformation to the host without visible influence to them. Contrary to informationhiding, steganalysis is a technology to detect hiding information. Currently, researchon steganalysis is mainly concentrated on mediums such as image, audio, video andtext. As for vector graphics, many efforts have been put on the information hidingmethods, while little attention has been paid to its steganalysis.In this thesis, a deep investigation is made into the steganalysis techniques forvector graphics. The characteristics of vector graphics, current research status ofimage and vector graphics’ hiding information is firstly presented, then the study ofthe steganalysis of image and analysis the shortcomings of the steganalysis forvector graphics are presented. After that, the relevant theoretical knowledge andtools, including probability theory and mathematical statistics, polar coordinatestransformation and OpenGL are introduced. Finally, two steganalysis methods areproposed, and their performance are also discussed in detail.(1) A steganalysis method for2D engineering graphics based on the statistic ofgeometric features is proposed. Firstly, obtain the number of2D engineeringgraphics’ strictly vertical and horizontal lines and identify the number of horizontalor vertical lines which are deviated from the straight line within a certain range, thenselect the ratio between the deviated lines and the normal lines as the statisticalcharacteristics, after that, a detection model is constructed based on the hypothesis.Experimental results show that the algorithm can detect hiding information in the2Dengineering graphics effectively.(2) A steganalysis method for three dimensional meshes based on the statisticsof geometric features is proposed. Firstly, the distance between each vertex and thecenter of gravity of three dimensional meshes is obtained; secondly, the distances aredivided into some distinct groups with equal range according to their magnitudes,after that, the distances in each group are mapped into the normalized range of [0,1];Finally, the variance of each group’ mean is calculated and a detection model isconsequence constructed based on the statistical characteristics. Experimental results show that the algorithm can detect hiding information in the threedimensional meshes effectively.The steganalysis methods proposed in this paper can expand the carrier field ofsteganalysis to a certain extent and can be well applied to vector graphics.