A comparative study of recent steganography techniques for multiple image formats

Published Online January 2019 in MECS

image is a combination of cover image plus hidden data.

This paper presents a survey of Steganography algorithms based on various image formats. The paper is organized as follows. Section I includes the introduction and technical properties of Steganography, applications, limitations, methods, and Steganalysis. Section II presents Steganography cover image formats, their methods/techniques and color model information of each image format. Section III describes the literature review of recent Steganography techniques. Section IV is devoted to comparative analysis. Finally, we wrap up the discussion in Section V.

Fig.1. An Example of Image Steganography.

• A.    Applications and Limitations

Steganography is very effective for hiding information and can be used for a number of applications like social, scientific and governmental applications. However, as always, every technology may also have some downsides. Steganography can also be misused for unlawful activities; some constraints are also encountered in using Steganography. Following Table 1 shows some applications, while Table 2 shows some limitations of Steganography.

Table 1. Applications of Steganography Techniques.

a.	Steganography is useful to transfer the secret message from source place to destination place.
b.	Steganography is also used to store and to transfer the information of secret location.
c.	Steganography can be used for secure online voting.
d.	It can be used for private banking.
e.	It can be used for the military purpose.

Table 2. Limitations of Steganography Methods.

a.	Terrorist for criminal activities can misuse it. To stop such illegal activities some governments have taken some corrective actions to restrict Steganography and the similar technologies. All these kind of technologies are under high surveillance.
b.	It can be misused by attackers to harm privacy concern for example in Film Industry (to plagiarise films), social media (to steal the personal information and pictures from WhatsApp Facebook Instagram etc.) Alternatively, software industry (for making pirated software).

• B.    Steganalysis

Steganalysis is an art of breaking Steganography method to expose the existence of secreted information. Fig. 2. shows the example of Steganalysis process. Steganalysis has two approaches, one is ‘specific Steganalysis' (specific for spatial domain or specific for JPEG) and the second one is ‘universal Steganalysis’ (for all types of image format). It will not go through the specific Steganalysis category over the Internet, because one cannot judge which type of format is being used by the transmitter. In specific Steganalysis approach, the embedding method is already known; whereas universal Steganalysis approach is not aware of any prior knowledge about the embedding method [32]. Steganalysis can also be used to measure the robustness of Steganography method. [30]. Several Steganalysis approaches are presented by researchers in [33 – 38, 42].

Fig.2. Illustration of the process of Steganalysis.

Steganalysis process is generally consists of six basic steps as shown in Fig. 3. Steganalysis senses suspected object over the Internet to break the Steganography method. The pre-processing step may apply image processing on the set of data images, for example, converting an image from color to greyscale or transformation or cropping or compression.

Steganalysis process also reduces dimensions of an image if required. The features should be rather different from the image without hidden message and for the stego-image. The larger the difference, the better the features are. The features should be as general as possible, i.e. they are effective to all different types of images and different data hiding schemes. Feature extraction, classifier design is another key issue for Steganalysis and the performance of a Steganalysis system. Combination of feature extraction and classifier design is evaluated by its classification success or error rate. [50]. Selection and design of the classifier are performed, based on extracted features. Steganalysis train the classifier according to the format required. In Steganalysis, classification is used to classify the set of the data object into the original data object and stego object. Some open source Steganalysis tools available are StegSecret, OpenPuff, StegDetect, StegBreak, StegSpy, Hiderman, Jsteg-shell, Jsteg-shell, JPhide and Seek, Camouflage, F5, Steganography Analyzer Real-Time, JPHide, JPegX, StegExpose.

• II.    Steganography Cover Image Formats

This section presents the discussion on various Steganography image file formats, their color models and different steganography methods / techniques used for various formats.

All the image formats consists of dissimilar characteristics, all contain different header information. The core difference between them is the amount of compression. For example, 24 bit RGB color image needs 9.6 megabytes storage if no compression is used. However, it requires much lesser space with compression. Finer details of image necessitates less compression, while more ratio of compression sacrifices the finer details. JPEG uses lossy compression, while Bitmap, PNG and TIFF images have lossless compression property.

• A.    JPEG (Joint Photographic Experts Group)

JPEG is one of the most commonly used image file formats. It is most widely used in digital cameras, memory cards, web pages, and image processing because JPEG format can compresses the image data into smaller file size and has low risk of attacks. JPEG uses lossy compression, which is a strong downside of it also. The error level is restricted to be below the perception threshold of human observer level. It does not allow editing and restoring images repeatedly, because more quality is lost every time you save an image in JPEG format. The signal delivered to the encoder is normally additive colours red, green and blue which are transformed into YCbCr components. JPEG uses the Huffman coder to encode the AC coefficients and differential encoding for the DC coefficients. It does not allow editing and restoring images repeatedly, because more quality is lost every time an image is saved in JPEG format.

which forms the part of the lossy level; and the second level is the Huffman coding that is a lossless data compression technique. JPEG image data embedding methods store secret data between these two phases [4]. DCT transformed cosine values cannot be back-calculated exactly and repeated calculation using limited precision number produces a rounding error hence, it is called lossy compression.

• B.    BMP (Bitmap)/RGB

BMP format offers compressed and uncompressed images file format in greyscale as well as in color mode. It also supports optional transparency. 8 bit Bitmap has a maximum of 256 colors per pixel. RGB is also available in 16 bits, 24 bits, 36 bits as well as 48 bits format. Here, 48 bits format images are considered as more color depth images as each channel uses 16 bits. In 24 bits format, each of the R, G and B channels use 8 bits and brightness intensity lies between 0 and 256. For 16 bits format, every pixel is two bytes and each color uses a precise number of bits. The syntax of Bitmap-File Structures [57] is as follows and details are as shown in Table 4.

BITMAPFILEHEADER

BITMAPINFOHEADER

RGBQUAD

BYTE bmfh;

bmih;

aColors[];

aBitmapBits[];

Table 3. Header Fields in JPEG File Structure [59].

JPEG header fields	Size
image_width	512
image_height	512
image_components	3
image_color_space	2
jpeg_components	3
jpeg_color_space	3
Comments	{}
coef_arrays	{1x3 cell}
quant_tables	{[8x8 double] [8x8 double]}
ac_huff_tables	[1x2 struct]
dc_huff_tables	[1x2 struct
optimize_coding	0
comp_info	[1x3 struct]
progressive_mode	0

A JPEG image consists of some coefficient matrices along with header information. A typical example of a JPEG image file structure has shown in Table 3 [59]. In the JPEG file structure, ‘Coef_arrays’ is one of the components in JPEG image file header. This component is a cell array of size 1 × 3 cell. We can divide each cell array into 8 × 8 blocks for easy and fast mathematical operations (less than 8 × 8 block does not contain enough information and greater than 8 × 8 blocks may not be supported by hardware or may take longer time too). Most of the information about the image lies in the DC coefficient which is the left top corner coefficient of DCT matrix. Other coefficients are known as AC coefficients. The JPEG coefficient values range from –1024 to +1023. Most of the AC coefficients have values of zero. JPEG compression has two levels: first DCT quantization,

Table 4. Bitmap File Structure.

Bitmap Structure Fields / Description	Contained Information
BITMAPFILEHEADER bmfh; [Bitmap file header]	It contains information about the field type, field size, and layout of a device. It is independent bitmap file.
BITMAPINFOHEADER bmih; [Bitmap information header.]	It specifies the dimensions, compression type, and color format for the bitmap.
RGBQUAD aColors [ ]; [Color table, and an array of bytes that defines the bitmap bits.]	The color table, defined as an array of RGBQUAD structures, contains all the basic color elements in bitmap. The number of bytes representing a scan line stored in the bitmap. The first byte in the array represents the pixels in the lower-left side corner of the bitmap and the last byte represents the pixels in the upper-right corner.
BYTE aBitmapBits[ ]; [The bitmap bits, consist of an array of BYTE values representing consecutive rows, or "scan lines," of the bitmap]	8-bit Bitmap contains the maximum number of 256 colors. Each pixel in the bitmap is denoted by a 1-byte index into the color table. 24-bit Bitmap has a maximum of 2^24 colors. The bitmap bmiColors member is NULL, and each 3-byte sequence in the bitmap array represents the relative intensities of red, green, and blue, respectively, for a pixel.

• C.    PNG (Portable Network Graphics)

PNG is used when we need a small file that maintains its original quality. It was designed especially for transferring images over the Internet. It supports a number of colors plus a varying degree of transparency. Transparency in the image allows an image to be moved or copied onto any other background image. PNG supports indexed color, grayscale and RGB. It supports palette-based images of 24-bit RGB or 32-bit RGBA colors, grayscale images, and full-colour non-palette-based RGB images. PNG is a lossless data compression. This means that all the data on the image is stored when the image is compressed, means there is no change in resolution. The PNG file always has first 8-byte signature values as shown Table 5 and four parts of chunks as shown in Table 6.

Table 5. PNG file with 8-byte Signature.

Field Values	Purpose Of Hexadecimal values
Hexadecimal 89	It has the high bit set to identify transmission systems, it do not support 8-bit data and to reduce the chance that a text file incorrectly interpreted as a PNG, or vice versa.
Hexadecimal 50- 4E - 47	It permitting an individual to identify the format without difficulty if it will viewed in a text editor.
Hexadecimal 0D - 0A	A DOS- style line ending to detect DOS Unix line ending conversion of the data.
Hexadecimal 1A	A byte that halts display of the file in DOS when the command type used the end-of-file character.
Hexadecimal 0A	A Unix-style line ending (LF) to detect Unix-DOS line ending conversion.

Table 6. Chunks within PNG.

Value Length	Chunk type	Chunk Data	CRC length
Four bytes	Four bytes	Length bytes	Four bytes

• D.    TIFF (Tagged Image File Format.)

TIFF format was developed in 1986 by an industry committee chaired by the Aldus Corporation. TIFF file extension is “.tiff" or ".TIFF". TIFF can handle a number of images within a single file. It is lossless format means it is an uncompressed file format when the image is compressed, there is no change in resolution. TIFF permit editing and resaving of the images without compression loss. TIFF offered options to use tags, layers, and transparency, and are compatible with photo manipulation programs like Photoshop. TIFF is the best choice if you need to edit the digital image. TIFF supports bi-level, grayscale, palette-color, and RGB full-color images.

• E.    Colour Models for Image Formats

A color model is a system for creating a whole range of colors from the basic colors. RGB and CMYK are the two common models used for image processing in Steganography. Overview of some more color models are given below.

• •    CMYK model

CMYK model (cyan- magenta- yellow -black). CMYK model uses the printing ink and here colors are the result of reflected light.

• •    RGB model

The RGB model uses light to display color. RGB color model consists of three basic colors red, green and blue. Light is added together in various combinations to reproduce a wide number of colors. The main purpose of the RGB color model is in the display of images on computer or TV. RGB model is an additive color model. Bitmap images used RGB model.

• •    HSV model

Hue means tint or tone, which is produced by "lightning", in terms of their shades of saturation and their brightness values. It is used in color editing software, but not in image analysis. Hue (h) color type ranges from zero to 360 degree, saturation color ranges from 0 to 100 % and value of brightness (v) ranges from 0 to 100 %. HSV and HSB model is same.

• •    HSL model

HSL, like HSV, is a 3-D representation of color. HSL stands also stands for hue, saturation, and lightness. The difference between the HSL and HSV model is: in HSL model the saturation and lightness components span the entire range of values.

• •    NCS model

The Natural Color System (NCS) is based on six colors that cannot be used to describe one another: white, black, red, yellow, green and blue; unlike RGB or CMYK model.

• •    Indexed colour

The color of each pixel is represented by a number. Each number called index corresponds to a color in the color table (the palette).

• •    Steganography Methods /Techniques

Some Steganography methods and techniques used for various image formats are described below.

• •    DCT (Discrete Cosine Transform) /

DWT (Discrete Wavelet Transform) Methods

DCT separates the image into 8*8 pixels blocks and embeds the secret bits by modifying the high or middle frequency. DWT divides the image pixel block into 4 sub-bands (LL, HL, LH, and HH), scan pixel from left to right horizontal manner and top to bottom vertical manner and then perform some addition and subtraction operations on pixels until the whole image get processed.

• •    Distortion Method

This method is used mostly on JPEG images. The secret bit is embedded using the distortion of the image and by calculating an error between original and stego image at the decoding stage in order to restore the hidden bits. The technique uses distortion functions and some error coding functions for Steganography.

• •    Spread Spectrum Method

Spread Spectrum radio transmission transmit messages below the noise level for any frequency level. This technique embeds secret bits in the noise and spreads secret data throughout the cover image. This technique can be merged with the error correction coding to ensure robust Steganography.

• •    Statistical Method

This method modifies the statistical property of an image for embedding. The cover image is divided into sub-images and one secret message bit is transmitted with a corresponding sub-image, transmitted in a way that the changes in statistical characteristics should not be visible.

• •    Adaptive Method

This method works for both spatial and transform domains. By using global statistical characteristics of the image, the method decides what changes can be done in the cover image, before processing the coefficients or pixels.

• •    LSB (Least Significant Bit Substitution) Method

Most common and popular method, in which LSB of a pixel is replaced by the secret message bit. Many modifications to the basic LSB substitution have been proposed, like the indicator method in [1].

• III.   Literature Review based on Different Image