Efficient construction of provably secure steganography under ordinary covert channels

Steganography is the science of hiding information within seemingly harmless messages or innocent media. This paper addresses the problems of efficient construction of secure steganography in ordinary covert channels. Without relying on any sampling assumption, we provide a general construction of secure steganography under computational indistinguishability. Our results show that unpredictability of mapping function in covertext sampler is indispensable for secure stegosystem on indistinguishability against adaptive chosen hiddentext attacks. We completely prove that computationally secure steganography can be constructed on pseudorandom function and unbiased sampling function under ordinary covert channels, that is, its security is inversely proportional to the sum of errors of these two functions, as well as the legth of hiddentexts. More importantly, our research is not dependent upon pseudorandom ciphertext assumption of cryptosystem or perfect sampling assumption. Hence, our results are practically useful for construction and analysis of secure stegosystems.