Speech steganalysis based on the delay vector variance method

This study investigates the use of delay vector variance-based features for steganalysis of recorded speech. Because data hidden within a speech signal distort the properties of the original speech signal, we designed a new audio steganalyzer that utilizes delay vector variance (DVV) features based on surrogate data in order to detect the existence of hidden data. The proposed DVV features are evaluated individually and together with other chaotic-type features. The performance of the proposed steganalyzer method is also discussed with a focus on the effect of different hiding capacities. The results of the study show that using the proposed DVV features alone or in cooperation with other features helps in designing a distinctive audio steganalyzer, as cooperation with other chaotic-type features provides higher performances for stego and cover objects.