An information-theoretic model for steganography
Information and Computation
The ultimate steganalysis benchmark?
Proceedings of the 9th workshop on Multimedia & security
The square root law of steganographic capacity
Proceedings of the 10th ACM workshop on Multimedia and security
Weighted Stego-Image Steganalysis for JPEG Covers
Information Hiding
Benchmarking for Steganography
Information Hiding
Fisher Information Determines Capacity of ε-Secure Steganography
Information Hiding
Estimating Steganographic Fisher Information in Real Images
Information Hiding
A general framework for structural steganalysis of LSB replacement
IH'05 Proceedings of the 7th international conference on Information Hiding
Steganalysis of Embedding in Two Least-Significant Bits
IEEE Transactions on Information Forensics and Security
Matrix embedding for large payloads
IEEE Transactions on Information Forensics and Security
Moving steganography and steganalysis from the laboratory into the real world
Proceedings of the first ACM workshop on Information hiding and multimedia security
| Hi-index | 0.00 |
We recently developed a new benchmark for steganography, underpinned by the square root law of capacity , called Steganographic Fisher Information (SFI). It is related to the multiplicative constant for the square root capacity rate and represents a truly information theoretic measure of asymptotic evidence. Given a very large corpus of covers from which the joint histograms can be estimated, an estimator for SFI was derived in [1], and certain aspects of embedding and detection were compared using this benchmark. In this paper we concentrate on the evidence presented by various spatial-domain embedding operations. We extend the technology of [1] in two ways, to convex combinations of arbitrary so-called independent embedding functions . We then apply the new techniques to estimate, in genuine sets of cover images, the spatial-domain stego noise shape which optimally trades evidence --- in terms of asymptotic KL divergence --- for capacity. The results suggest that smallest embedding changes are optimal for cover images not exhibiting much noise, and also for cover images with significant saturation, but in noisy images it is superior to embed with more stego noise in fewer locations.