Instead of using correlation, if we allow the use of the original document, then one can calculate the correlation between the difference signal and the watermark, i.e. which is more reliable. This is the method described in Slide30.
What is steganography? What are the design objectives for a steganography system?
Steganography refers to the technique that embeds a secrete message in a cover media. The design objectives include: 1) the stegoed media (the cover media after embedding) should still look like a typical, normal document; 2) it should be difficult for a steganalysis system to tell the presence of a secrete message from the stegoed media. 3) the intended recipient of the stegoed media should be able to detect and decipher the embedded message reliably.
Propose one possible technique, to embed a secret message in an image or a song or a video, and a corresponding algorithm for deciphering the message. Remember that the image (or song or video) after embedding (the stegoed media) should look indistinguishable from the original one, and the intended recipient of the stegoed media should be able to decipher the secret message (with the knowledge of a secret key and stego algorithm) exactly, and that it should be difficult for a steganalysis system (the Warden Wendy) to detect the presence of the secret message.
A simple technique is to change the least significant bits of randomly chosen pixels (based on a secret key) according to the secrete message to be embedded. The secrete message is first encrypted and converted to a binary sequence of 0 or 1s. Then the LSBs of chosen pixels are reset to the bits corresponding to the secrete messages. The recipient can simply retrieve the message by collecting the LSBs of the same set of pixels (based on a key). Because only the LSBs are changed, the modified image should look very much like the original image.
What is steganalysis? What are the design objectives for a steganalysis system? What are the two types of staganalysis techniques?
Steganalysis refers to the detection of presence of embedded data in a medium. The design objective is to maximize the “hit” rate, while minimize the “miss” rate, and in the mean time to minimize the false detection rate.
One may develop staganalysis techniques for a given known embedding method. Such methods tend to have higher detection rate, but is useless for a new embedding method. Universal steganalysis technique, on the other hand, aims to detect the presence of embedded data without knowing the embedding technique. This is possible as the embedded data are likely to change the statistics of the original media, no matter what embedding algorithms are used.