High Performance Steganographic Scheme Applying Time-Varying Convolutional Embedding Codes

A matrix embedding code was developed as a commonly used steganographic technique in which a parity-check matrix is used to perform embedding. However, a drawback of high decoding complexity for linear block codes by using the maximum-likelihood algorithm is unrealistic. This paper proposes a simple and effective trellis embedding scheme for binary messages. Compared with a matrix embedding algorithm that uses linear block codes, the proposed scheme is more appropriate for embedding messages in the case of linear block codes with a long length. The proposed algorithm uses time-varying convolutional codes as the embedding method and yields a favorable structure of time-varying convolutional codes for steganography. The proposed method employs maximumlikelihood decoding based on trellis construction to identify the coset leader of convolutional codes for large payloads. The experimental results show that the embedding efficiency of the proposed scheme is substantially superior to that of the scheme using linear block codes. Key-Words: Steganography, matrix embedding, embedding efficiency, linear block code, convolutional codes