Comparative study on the evaluation and benchmarking information hiding approaches based multi-measurement analysis using TOPSIS method with different normalisation, separation and context techniques

Abstract

This paper presents a comparative study of evaluation and benchmarking information hiding approaches based on multi-measurement analysis using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method in terms of different parameter settings because several settings of TOPSIS are continually making complex decisions within any study case. The problem is extended from original TOPSIS in terms of different contexts, distance and normalisation techniques. However, each technique changes the behaviour of data differently. The literature presents no evidence that any of those techniques affects the final decision. Supporting results are needed to justify which technique is the best for this particular study on evaluation and benchmarking digital watermarking approaches. For these purposes, two experiments are performed. In the first experiment, a noise-gate-based digital watermarking approach is developed, and a scheme for noise gate digital watermarking approach is enhanced. A total of 60 audio samples with different audio styles are tested using the two algorithms. A total of 120 samples are evaluated according to three different metrics, namely, quality, payload and complexity, to generate a set of digital watermarking samples. In the second experiment, three different weights are selected in the decision making process. For the first weight, 50% of importance is given to size and 25% each for quality and complexity. In the second weight, 50% is given to quality and 25% each to size and complexity. In the last weight, 50% of importance is given to complexity and 25% each to size and quality. In the decision making solution, the algorithm adjustment and parameter selection of TOPSIS are identified and tested based on three parameters, namely, normalisation, separation and contexts. In the normalisation process, two different linear normalisation and vector normalisation techniques are used. In the separation process, two different techniques, namely, Manhattan and Euclidean distances, are used, and each of these measurements has different philosophies. In the context process, two different contexts, such as decision making based on signal and group, are used. The results are as follows: (1) The preferred normalisation technique is followed by vector normalisation. (2) We recommend linear normalisation with TOPSIS by integrating the results from our experiment and the literature experiment. (3) The separation measurement has an effect on decision making algorithms, and the Euclidean measurement in the separation process reflects more accurate results than the Manhattan measurement. (4) The framework selects one of the two context settings, that is, group or individual, which will give the same results. However, group decision making is the recommended approach in the case of selection where priority weights are generated from the evaluators.