Steganographic systems use Syndrome Trellis Coding (STC) to control the selection of embedding positions in a cover, subject to a set of stochastic constraints. This paper reports observations from a series of experiments on the ability of Syndrome Trellis Coding to approximate independent Bernoulli random variables. We find that approximation errors are generally small except for some outliers at boundary positions. Bivariate dependencies between embedding changes do reveal the use of the code and its parameters. While risky outliers can be hidden by permuting the cover before coding, or avoided by using the proposed “outlier corrected” variant OC-STC, the aggregate bivariate statistics are invariant to permutations and therefore constitute a potential security risk in the presence of powerful attackers.
On the statistical properties of syndrome trellis coding
Cecilia Pasquini;
2017-01-01
Abstract
Steganographic systems use Syndrome Trellis Coding (STC) to control the selection of embedding positions in a cover, subject to a set of stochastic constraints. This paper reports observations from a series of experiments on the ability of Syndrome Trellis Coding to approximate independent Bernoulli random variables. We find that approximation errors are generally small except for some outliers at boundary positions. Bivariate dependencies between embedding changes do reveal the use of the code and its parameters. While risky outliers can be hidden by permuting the cover before coding, or avoided by using the proposed “outlier corrected” variant OC-STC, the aggregate bivariate statistics are invariant to permutations and therefore constitute a potential security risk in the presence of powerful attackers.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
