Research On Steganography In JPEG Images

Information hiding is the technology of embedding secret messages into cover images using redundancy of the digital signal. Once the messages have been embedded, it may be transferred in public. In order to communicate secretly, both the sender/receiver and the existence of communication process are hided. Therefore, the illegal user can not judge which image contains the secret information from the abundant web resources.This paper conducts the following works on the basic theories of information hiding technology and steganography in images:(1) Conduct different pretreatments according to different sorts of the secret messages. When the secret data is text, encrypt it before embedding, else if there is an image to be embedded, then use scramble methods on it.(2) A steganographic method based on modulus arithmetic is proposed, which is conducted on the quantized DCT coefficients of YUV. Three coefficients at the same frequency in YUV components are selected to compose a triplet as the message carrier, and then the embedding capacity in each triplet is decided according to its property. The modulus 2 or modulus 4 arithmetic operations are applied to the valid triplet to embed 1-bit or 2-bit secret messages. When modification is required for data embedding, a random number is produced to decide the flipping direction that approximately preserves the histogram of quantized DCT coefficients. The improved shortest route modification scheme (I_SRM) is applied to reduce the distortion of the cover image.(3) In JPEG images, a novel steganographic method in DCT domain based on a chaos system and genetic algorithm is presented. The main idea of this method is to minimize the degradation of the stego image by finding a best mapping function between secret message and host image at global scope.Simulation results show that these two steganographic methods have high embedding capacity and good invisibility, meanwhile, are robust against common steganalysis. Moreover, the process of messages extraction is simple.