A comprehensive evaluation framework for EMD-based steganography: balancing capacity and security with upper bound analysis

A comprehensive evaluation framework for EMD-based steganography: balancing capacity and security with upper bound analysis

연구 분야: Strategies

학회: The Journal of Supercomputing

Steganography is a technique to hide the presence of secret communication and can be used when one of the communication elements is under the enemy’s influence. The primary measure to evaluate steganography methods in a specific capacity is security. Therefore, in a certain capacity, reducing the number of changes in the cover media leads to a higher embedding efficiency and, thus, higher security of a steganography method. Generally, security and capacity conflict and the increase of one lead to the decrease of the other. A single criterion representing security and capacity simultaneously can help compare steganography methods. Exploiting modification direction (EMD) and methods based on it are a type of steganography techniques that optimize the number of changes resulting from embedding (security). Despite their effectiveness, existing evaluation metrics for EMD-based methods lack precision and comprehensiveness. The present study aims to provide an evaluation criterion for this group of steganography methods. In this study, after a general review and comparison of EMD-based steganography techniques, a method is presented for their precise comparison from the perspective of embedding efficiency. Initially, we conduct a thorough review and comparative analysis of existing EMD-based steganography methods to identify their strengths and limitations. Building on this foundation, we introduce an enhanced embedding efficiency formula that accurately quantifies the impact of one or more-pixel changes, providing a more nuanced assessment of embedding performance compared to traditional metrics. Our results demonstrate that the proposed embedding efficiency formula offers superior performance evaluation, particularly in scenarios involving multiple pixel alterations. Furthermore, we establish an upper bound analysis to determine the theoretical maximum embedding efficiency achievable for any given capacity. This upper bound serves as a benchmark for assessing the optimal performance of EMD-based methods. Finally, leveraging the upper bound, we present an additional evaluation criterion that facilitates a more precise and meaningful comparison of EMD-based steganography methods.