Article Preview
TopIntroduction
Information hiding is a technique to hide secret information in other irrelevant carriers to make embedded information invisible. Traditional information hiding usually modifies carrier redundancy for embedding. However, it is challenging to escape detections from steganalysis for it always leaves modification traces. To address it, CIH (coverless information hiding) is proposed. Unlike modification-based information hiding, CIH generates or searches stego carriers according to secret information directly. In CIH, SCIH (search-based CIH) is a typical class of strategies. The main idea of SCIH is finding appropriate carriers from a big database and using keywords following location tags to express secret information.
According to different properties of text and image, SCIH has two categories: one is SCIHT (SCIH for text), the other is SCIHI(SCIH for image).
SCIHT usually regarded text features as location tags to mark positions of secrets such as radicals of Chinese characters, hash values about series of Chinese character, ranks of English words, and so on. For example, Chen et al. (2015) regarded Chinese character radicals appeared in the top 50 as location tags and used Chinese characters or words following location tags as keywords. To avoid the extraction confusion in the work of Chen et al. (2015), Zhou et al. (2016b) restricted the range of location tags to the first appearing top 50 radicals in texts, and the hidden information following location tags must be a single character. Chen et al. (2017) further transformed a series of Chinese characters into -bit hash value as location tags to extract secret Chinese words following them. To enhance the hidden capacity of the work of Chen et al. (2017), Chen et al. (2018) regarded the high-frequency combination of Chinese word following location tags as keywords. To reduce the number of texts in database and improve the success rate of secret matching, Xia et al. (2017) used 12 predefined positions as location tags in one text, regarded LSBs of located Chinese characters as keywords. Based on English linguistic characteristics, Zhang et al. (2017a) and Zhang et al. (2017b) employed both the word rank map and the frequent word distance to express secret information.