A New Approach Based on Multi-Dimensional Evaluation and Benchmarking for Data Hiding Techniques

This paper presents a new approach based on multi-dimensional evaluation and benchmarking for data hiding techniques, i.e., watermarking and steganography. The novelty claim is the use of evaluation matrix (EM) for performance evaluation of data hiding techniques; however, one major problem with performance evaluation of data hiding techniques is to find reasonable thresholds for performance metrics and the trade-off among them in different data hiding applications. Two experiments are conducted. The first experiment included LSB techniques (eight approaches) based on different payload results and the noise gate approach; a total of nine approaches were used. Five audio samples with different audio styles are tested using each of the nine approaches and considering three evaluation criteria, namely, complexity, payload, and quality, to generate watermarked samples. The second experiment involves the use of various decision-making techniques simple additive weighting (SAW), multiplicative exponential weighting (MEW), hierarchical adaptive weighting (HAW), technique for order of preference by similarity to ideal solution (TOPSIS), weighted sum model (WSM) and weighted product method (WPM) to benchmark the results of the first experiment. Mean, standard deviation (STD), and paired sample t-test are then performed to compare the correlations among different techniques on the basis of ranking results. The findings are as follows: (1) A statistically significant difference is observed among the ranking results of each multi-criteria decision-making (MCDM) technique, (2) TOPSIS-Euclidean is the best technique to solve the benchmarking problem among digital watermarking techniques. (3) Among the decision-making techniques, WSM has the lowest rank in terms of solving the benchmarking problem. (4) Under different circumstances, the noise gate watermarking approach performs better than LSB algorithms.