An Analytical Framework for the Optimal Design and Detection of Covert Channel Communications
Abstract
We propose an iterative generalized least squares procedure to recover unknown messages hidden in image hosts via spread-spectrum embedding. Neither the original host nor the embedding signature is assumed available. We demonstrate that for hidden messages of sufficient length (data sample support), recovery can be achieved with probability of error close to what may be attained with known embedding signature and known original host autocorrelation matrix. For small hidden messages, the signature estimate calculated by the iterative generalized least squares procedure can be fed as initial value to a (computationally costly) expectation-maximization signature identification scheme that we derive. Message recovery can again be carried out successfully by means of a linear sample-matrix-inversion minimum-mean-square-error receiver.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 2006
- Accession Number
- ADA451407
Entities
People
- Dimitris A. Pados
- Michael J. Medley
- Stella N. Batalama
Organizations
- State University of New York at Albany