Article Preview
Top1. Introduction
DNA, as we all know, is the genetic codes for human beings and other plants and animals. A DNA sequence is a string of 4 nitrogenous bases: adenine (A), thymine (T), guanine (G), and cytosine (C). Most of current discovered DNA sequences are stored in GenBank of NCBI and other DNA databases. Normally, a sequence file in GenBack consists of two parts: the original sequence and the annotation. The annotation includes several labels such as LOCUS, DEFINITION, ACCESSION, SOURCE and FEATURES, which help to describe the original sequence. These annotations usually occupy a lot of the storage space of the database. Besides the storage problem, once the sequence is tampered by evil attackers, researches based on those tampered sequences will result in disastrous consequences. Thus, a method that could either solve the storage problem or detect and restore tampers on DNA sequence is of much help.
Reversible data hiding is a widely used technique in digital media field for annotation, watermarking and authentication. As its name indicates, reversible data hiding losslessly hides data into different media. When these data are extracted, the original media are perfectly retrieved. With different types of data embedded, reversible data hiding can be applied to different applications. For example, embedding media-related annotations could make such media self-describable. This is extensively used in medical image processing. In addition, if the content reference information is embedded, we say that the medium is watermarked. In general, there are three types of watermark: fragile watermark, semi-fragile watermark, and robust watermark. Fragile watermark fails to be detectable so it is widely used for tamper detection. Semi-fragile watermark is designed to resist benign transformations in order to detect malignant transformation. A robust watermark can tolerate a designated transformation and therefore, copy protection applications always choose this kind of watermark.
Introducing the traditionally used reversible data hiding techniques to DNA sequences will help to solve the storage problem and to authenticate the sequences. Some scholars have proposed their data hiding and watermarking methods. Shimanovsky et al. (Shimanovsky, 2003) firstly proposed a method in which utilized the redundancy of codon to amino acid mapping. There are 43 = 64 different codon combinations with three Nucleotides. But all possible amino acids and signals are 22. So a codon in a sequence can be substituted by its peer codons that will be translated into the same amino acid as the original one does. Hence, several bits can be embedded into each codon. In (Chang, 2007), Chang et al. provide two methods One of them is based on lossless compressing techniques. They losslessly compressed the DNA sequence and then append the secret data to the end of the compressing result. Another one adopted the difference expansion way to hide data, which is widely used in digital image data hiding. Another three methods: the insertion method, the complementary pair method and the substitution method were proposed by Shiu in (Shiu, 2010). Unfortunately, however, in order to restore the original content, all three methods have to make sure that a reference DNA sequence is transmitted to the client. With this premise, all three methods cannot be denoted as reversible data hiding methods. After that, Guo et al. (Guo, 2012) improved the substitution method. But the reference DNA sequence is still needed.
In this paper, we propose a new reversible data hiding method on DNA sequences. In this method, we utilized degenerate base symbols as labels to mark DNA sequences. Hence no reference DNA sequences are needed. Based on this method, two applications are proposed. One annotates DNA sequences by hiding GenBank sequence annotations to make sequences self-describable. The other is a fragile watermark scheme, in which the watermark to be embedded can be used to detect tampered area(s) in DNA sequences. Even further, if tampered area(s) takes up only little portion of the entire sequence, the original content can be exactly retrieved.