2024 Second International Conference on Advance in Information Technology (ICAIT-2024)

A Study and Analysis of Reversible Data Hiding

Techniques
Yogish Naik G.R Namitha R Shetty Vidyasagar K.B
Department of Computer Science Department of Computer Science Department of Computer Science
Kuvempu University Kuvempu University Kuvempu University
Shivamogga,India Shivamogga,India Shivamogga,India
[email protected] [email protected] [email protected]

Abstract—This overview paper explains several kinds of unique type of data concealing methodology known as RDH.
Reversible Data Hiding (RDH) methods. Reversible data Reversible steganography once hidden data is retrieved, can
hiding involves concealing data within host media, such as recover the initial carrier with negligible distortion or without
audio, video, and image files. RDH, or lossless data hiding, any distortion at all[4]. The image quality distortion should
makes it possible to retrieve the embedded information and be very less after information embedding From the
then restore the original cover without sacrificing any perspective of information concealing, reversible embedding
information. Numerous RDH approaches have been developed. conceals information within a digital image so that it can be
These techniques have been reviewed and summarized in this
decoded by a legitimate person and the image can be
paper. Different approaches like LSB modification, Prediction
error method, Difference expansion and Histogram shifting
returned to its original state.
methods have been reviewed. Encryption is used in images to II. RELATED WORK
provide security. Each RDH approach has advantages of its
own, and no single strategy is ideal or suitable in every Rathika R et al.[4] has proposed a technique on
situation. This paper offers a thorough analysis of all the Reversible Data Hiding. Here it presents an in depth survey
primary approaches of RDH that have been used, as on numerous techniques obtainable for RDH. That contains
documented in the paper. Histogram shifting, LSB modification method, Difference
Expansion, Interpolation Technique. All these methods
Keywords—Reversible Data Hiding, Histogram Shifting, directs at replicating the initial image throughout that the
Difference Expansion, Prediction Error. information was concealed with most accuracy. Reversible
Information Stego images are restored to their perfect states
I. INTRODUCTION
employing hiding techniques after hidden information has
Data is embedded into digital media as part of the data been extracted. These methods is classified into 3 groups that
concealing process for security reasons. Data concealing is is supported data compression, supported pixel value DE,
the art and science of transmitting confidential information in supported histogram shifting. There are numerous techniques
a suitable multimedia carrier, such as an image, audio, or which are obtainable for the RDH approach. One of those
video file. Data hiding consists of watermarking and preferred approaches is the DE methodology, which
steganography techniques. RDH is a process in which the computes and enlarges pixel group differences to
data is concealed within the host media, maybe as a cover accommodate additional bits. The histogram shift approach
image[1]. RDH approach can, when the data has been modifies the pixel grey values subtly by using the zero and
extracted, restore the initial image without causing any loss peak points within the histogram image to incorporate
of information. RDH, also known as lossless data information into the picture. RDH method is Pair-Wise
embedding, is an approach that allows unseen data, or a Logical Computation (PWLC) for binary images. Every
payload, to be reversibly embedded into a digital image[2]. methodology has a number of benefits in addition to its own
The image's quality degradation during data embedding downsides. After retrieving the encoded information, a
must be minimal. Reversible data embedding has an revolutionary reversible data concealing approach allows the
interesting feature, the initial image can be restored by precise retrieval of the actual image. The data embedding
removing the embedded data. Integrity authentication, media approach called LSB modification is suggested, this adds
notation, copyright protection, and other uses can all benefit new operational points to the capacity distortion (C-D) curve.
from the approach known as data hiding, which embeds In order to fit the payload information, it adjusts the host
information into covers for different media files. The image's lowest levels rather than bit planes. A prediction-
majority of data concealing techniques achieve perceptual based entropy conditional coder uses the host's fixed sections
transparency by embedding messages into the cover media, as side-information. Bits of information are embedded by
such as an image or video, and only changing the least adjusting specific host picture properties. Both the capacity
important portion of the cover to create the marked media[3]. for lossless data embedding and compression efficiency will
Usually, the embedding procedure causes the cover to increase.
become permanently distorted, making it impossible to Juhi Gupta et al. [2] has proposed a location map scheme
recreate the original cover from the marked cover. for reversible data hiding using histogram shifting, here it
In applications like law forensics, the military, and presents a systematic location map theme by employing only
medical imaging the deterioration of the original cover is not one bit, which could improve image recovery. The embedded
permitted. To recover the hidden message and recover the secret information can’t be retrieved once the information of
initial cover without any loss in these situations, we employ a peak and zero point of histogram aren’t transferred to the

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

2024 Second International Conference on Advance in Information Technology (ICAIT-2024)

receiver. The location map is to provide a solution to resolve using an improved overflow placement map. The histogram
once the chosen minimum point isn’t zero and to store the will be divided into two parts in this embedding theme: an
reversible information. Here it presents a review of RDH inner region for embedding and an outside part for shifting.
approach which have higher information concealing Low distortion transform makes use of the two
capability and invisible artifacts is represented. transformation types to incorporate important information
into pictures. Incorporate the data not just into the current
In order to use the reverse technique operation from the
pixel but also into its anticipated content. This technique
stego-image inside the receiver, the original picture is often
produced less distorted images.
recovered. It is evident that the NSAS theme's information
embedding capacity can determine the peak point's grayscale III. REVERSIBLE DATA HIDING TECHNIQUES
value. The peak point in the marked image, however, differs
from the initial image because when secret information is The data-concealing approaches fulfilling this
included, the peak point in the stego-image is altered. The requirement are invertible or RDH systems. RDH comprises
recipient must get the grayscale values of the highest and information embedding into media in a method that considers
zero point inside the selected image in order to extract the both the original media and the retrieval of the inserted
embedded secret information. information. RDH joins different methods to guarantee the
reversibility. Its practicality is principally because of the
In NSAS, the embedded confidential information can’t be lossless compressibility of regular pictures.
extracted once the peak and zero point of histogram don’t
seem to be send out to the receiver. The primary drawback in RDH is an approach in which the information is
this case was the method of transmitting the added concealed within the cover media, which could be a
information, the histogram's highest and zero points, from the image[7]. Reversible data concealing is an approach that,
sender to the recipient. HKC algorithm was proposed, a once the information has been recovered from the cover
sturdy reversible data concealment theme supported image, allows for the lossless recovery of the original image.
histogram shifting technique, to enhance the on top of By adopting a method called lossless data embedding,
disadvantage. The location map is utilised to safeguard the confidential data can be reversibly included into a digital
reversible information and provides a way out once the image. The quality degradation of the host image should be
chosen minimum purpose is not zero. Moreover, resolution minimal. The RDH achieves reversibility once the embedded
here wastes the storage amount and decreases the initial data data is restored. Usually, the data embedding process causes
concealment capability. An effective location map that the cover medium to permanently lose some of its properties.
reinforces the data hiding capability by reporting changes to But in some applications the host image or cover image is
the selected minimum purpose using just one bit, which can as important as that of embedded data. Applications like
enhance the data concealment schemes has proposed. medical field, defense and in law forensics the restoration of
M. Jeni et al. [6] presented a RDH inside the videos cover is also important.
by utilizing low distortion transform. Here, reversible RDH is an approach that modifies the pixel values of
information hiding is employed to provide a novel way of digital images to insert data for covert communication. Once
concealing important information in videos. Removing the the hidden data has been eliminated, the cover image can be
valuable information from a video sequence, combining the restored to its initial state. Several techniques[2] have been
images features to create videos, and then inserting the most used in RDH to restore the cover image. In this paper
important message segments into the videos. To protect and different RDH methods are analyzed. Fig. 1. represents the
recover the data, two different transformation types are used, structure of RDH which demonstrates about secret key,
which greatly increases security. The reason this embedding embedding and extraction process in RDH.
procedure produces minimal distortion films compared to
standard techniques is because it breaks the important
information into all four selected pixels instead of embedding
it into the current pixel. It ensures that the Mean Square
Error is as small as possible.
Using DE of vectors, Difference expansion extends
generalized integer transform. Cross-color and Recursive
embedding they employ two different types of embedding.
By doing this, the computational efficiency and hiding
capacity will both rise. Sorting and Wavelet techniques
utilizes 1 technique for embedding information within Fig. 1. Structure of Reversible Data Hiding.
pictures specifically LSB. By utilizing the information that is
included within the most vital bit plane LSB predicts the A. LSB Modification
least significant bit plane . The LSB [8] is the broadly utilized spatial domain
information concealing system. It implants the same measure
To form a good methodology for information embedding, of information as the LSB pixels. LSB is an approach to hide
DE Embedding method goes to merge the histogram shifting secret information in the LSB’s of the pixels without
and prediction-error expansion. The location map has presenting numerous noticeable distortions and is one of the
stopped employing histogram shifting. obtaining the PE from easiest strategies. Changes in the pixel values in this method
a pixel's surrounding area and including the data bit into the are intangible for human eyes. This substitution approach
expanded PE. DE can increase the sponginess of the same by substitutes a bit from the secret information for the rightmost

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

2024 Second International Conference on Advance in Information Technology (ICAIT-2024)

bit in the binary representation. This method replaces the  Use the inverse-integer-transform to reconstruct the
LSB with secret information bit. image that was retrieved.
LSB insertion is a basic methodology for implanting data C. Prediction Error Using Expansion
in a cover media[14]. The last bit in the binary representation This method is one among RDH method which is based
of some or all the bytes of an image are replaced by the on the prediction error[11]. This uses the predicted system
secret information bit. A bit of each R-G-B components can for embedding process. There are different types of
be used in case of 3 byte image representation. For instance, predictors like vertical, horizontal, casual, casual-weighted-
a matrix for 3 pixels comprising of a 24 bit image can be as average and SVF. The predictor that is widely used is the
per the following: median-edge-detection.
00100101 00011100 11101001 An effective data hiding approach based on the predict
00100110 00101111 11010111 error method is given to improve the embedding capacity.
Reversible data concealing based on prediction error use
01101100 00100010 11001011 predicted systems to embed data, several predictors have
Suppose the number is 400 then the equivalent binary been created. They are SVF, Causal weighted average,
representation is 110010000. The subsequent lattice will be vertical predictor, horizontal predictor, and Causal. The
shown as below. median edge detection (MED) predictor is one popular
predictor. the embedding methodology as follows:
00100101 00011101 11101000
 Step 1: Use three by three blocks to divide the cover
00100110 00101111 11010110 image.
01101100 00100010 11001010  Step 2: Determine the prediction error value between
So if the number was inserted into the initial 9 bytes of the surrounding pixels and its block’s center by using
the lattice just the 4bold bits should have been changed by the block's center as the base point.
implanted message. Taking the average only half of the bits  Step 3: Are they able to forecast the error value
should be adjusted to conceal a message. histogram and identify the peak
B. Data Embedding Using Difference Expansion  Step 4: Integrate the encrypted information into the
The difference expansion method [9] is the best RDH peak point's two-side region.
method which provides less distortion and more embedding
capacity and low distortion. Here the adjacent pixel  Step 5. obtain the stego-image.
differences are calculated. Only those differences are chosen Use the below mentioned steps to retrieve the
that do not result in issues with overflow or underflow. Then confidential information:
the message confirmation code, original substance rebuilding
data and extra information are embedded in those pixels.  Step 1: Use 3 * 3 blocks to divide the stego-image.
The steps in embedding are:  Step 2: Determine the prediction error value between
the surrounding pixels and its block's center by using
 Calculate the differences of adjacent pixels. the block's center as the base point.
 Determine the variable differences in those  Step 3: Locate the secret data by locating the highest
distinctions. point on the predict error value histogram.
 Few are selected to be expandable by 1-bit which  Step 4: Retrieve the original image and restore the
builds the alterable bits. error histogram. The slight variation between adjacent
 Create the location map which has location data of the pixels is the main feature of the prediction error
selected pixels. approach. Since the statistical data focuses on the zero
point, the embedding capacity will be increased and
 Gather primary LSB values. the image quality will be maintained.
 Substitute data inside the location map. The examples are shown below.
The steps in extraction are: Horizontal:
 Calculate the differences of adjacent pixels. p1 (x, y) = p(x-1, y) (1)
 Determine the variable differences in those Vertical:
distinctions.
p1 (x, y) = p(x, y-1) (2)
 Gather the Least significant bits of difference values. Causal-weighted-average:
 Separate the packed original variable bit stream and p1 (x, y) = (p(x-1, y) + 2p(x, y-1) + 2p(x-1, y) + p(x-1,
decipher the location map. y+1)/6 (3)
 Decompress the packed isolated bit streams and
rebuild the first image supplanting the variable bits.
1) Prediction Using the center pixel :

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

2024 Second International Conference on Advance in Information Technology (ICAIT-2024)

Here the data is split into non-overlying structure, so this Fig. 2.-(b), where we can observe that the number nine is ten,
method is extremely valuable[4]. Here the center pixel is the zero point is seven, and the peak point is nine. Thus, 10
more important as this can be used to predict the pixels. As bits can be embedded into the host picture. The imbedded
the other pixels have the same values as the center pixel. The data, we think, is 2 (1101110001). Stego-image and
distinction between the original pixel worth and the worth recovered image are represented in Fig. 2.-(a) and 2.-(b),
that is guessed is named the error-block. Table I depicts the respectively.
prediction using central pixel.

TABLE I PREDICTION USING CENTRAL PIXEL

In the Table I, the pixel Pa, Pb, Pc, Pd, Pe, Pf and Pg are
predicted as P. The resultant error-block is shown in Table II.

TABLE II PREDICTION ERROR BLOCK

Fig. 2. (a) Cover Image (b) Histogram of Image (c) Stego Image
D. Histogram Shifting Expansion (d) Recovered Image

The RDH approach that rely on histogram-shifting needs

E. Vector Quantization Based Technique
to embed information by moving the histogram in a fixed
region[10]. Additionally, the peak and zero point are two RDH based on vector quantization (VQ) is an approach
crucial points in these systems. The value in grayscale in the compression domain that is distinct from the first four
represents the highest number of pixels in the image's schemes mentioned above[1]. VQ provides a effective
histogram, which is represented by the highest point. The compression method, which is extensively utilized due to its
zero point is typically the place on the histogram where the higher efficiency and convenient implementation. To embed
number is 0. To maximize the embedded capacity, the zero data, utilize the left and top blocks to forecast the present
point is chosen to be the smallest amount of pixels. Utilizing block. To encode the current block adjacent blocks are
approaches based on histogram-shifting, a piece of the secret utilized during the encoding phase, however extra flag bits
message was transported utilizing the pixel at the peak point, are necessary. VQ provides a lossy compression method.
whereas no hidden data was embedded and the other pixels Rebuilding the signal with some loss results in a reduced data
were altered. During the embedding procedure, the pixels rate for fewer saves in photos, videos, and transfers. Fig. 3.
between the highest and zero pairings were changed. The demonstrates the vector quantization.
histogram shifting based data concealment method's Code words are taken from the codebook and mixed with
concealing capability is determined by the capability of a collection of representative samples. Fig. 3. displays the
pixels in highest points, the more pixels in the greatest point, coding book's flow chart. In the process of decoding, the sets
greater the hiding capability. of index are searched using simply the output of the coding
Using more highest points and zero pairs might maximize process. Then, locate the matching code words in the
the hiding capacity[11]. It can be challenging to identify codebook that correspond to the original sorting and retrieve
further peak-zero point pairings since the zero points are the image data. The decoding procedure is complete after all
difficult to search[12]. The histogram shifting algorithm indexes are processed.
procedure contains the following procedure: 1. The image’s By using data training, this method generates a codebook
histogram is created at first. 2. Next step is to determine the that combines each representative vector with a code word to
highest and zero points. 3. Assuming peak point as x and represent it symbolically. The codebook's size and domain
zero point as y where x>y. the points are shifted between determine the compression rate. The VQ encoding codebook
y+1and x-1. 4. The peak point is reserved if the embedded bit includes the generation, optimization, encoding, and
is 1; if not, reduce 1 to alter the peak point value. 5. It is decoding. First, choose the original codebook that causes the
necessary to record and embed the pixel locations in the local optimization issue after optimization is crucial for
minimum point in order to meet the requirements of producing. To increase the codebook Linde-Buzo-Gray
reversibility[13]. Next, note the zero points, the peak point, (LBG) was utilized often. The images are separated into
and any further details. equal-sized, non-overlapping blocks. Every block has the
An example of the fundamental steps of the histogram- same dimensions when it comes to coding and is treated as a
shifting based on RDH is provided in figure. We take Fig. 2.- vector. Code words are taken from the codebook and mixed
(a) to be the cover image. The image's histogram is shown in with a collection of representative samples.

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

2024 Second International Conference on Advance in Information Technology (ICAIT-2024)
180000
160000
140000
120000
100000

Capacity
80000
60000
40000
20000
0
LSB DE PE HS VQ
RDH Methods

Fig. 5. Comparison of Capacity for Different Approaches

Fig. 3. An example of VQ encoder
V. CONCLUSION
IV. PERFORMANCE ANALYSIS OF RDH TECHNIQUES The increasing popularity of RDH approaches can be
attributed to the carrier medium reversibility at the receiving
Sl. Method Authors PSNR Capacity end following the retrieval of secret information. After
No Used analysing the various reversible data hiding algorithms, it can
1 LSB Mohammadi, A 41.1 9325
Modification et al 7
be said that the RDH approaches mentioned are a useful tool
for data transport that is of secure. Each approach has pros
2 Difference Osamah M et 44.2 53628 and cons of its own.
Expansion al 0
3 Prediction Bini A.A et al 45.11 86178 REFERENCES
error [1] Manisha G et al. “Reversible Data Hiding Technique and its Type, a
4 Histogram Sahu, M et al 48.9 163840 survey”. IOSR Journal of Computer Engineering (IOSR-JCE), PP 43-
Shifting 3 48, 2016.
[2] Zhicheng Ni et al. “Reversible data hiding. in IEEE Transactions on
5 Vector Manisha G et 35.3 39321 Circuits and Systems for Video Technology”, vol. 16, no. 3, pp. 354-
quantization al 9 362, 2006.
[3] Jitha Raj.T et al. “A Survey Paper on Various Reversible Data Hiding
Table III demonstrates the performance analysis of Techniques in Encrypted Images”, IEEE International Advance
different RDH techniques and the comparison of different Computing Conference (IACC), PP 1139-1143, 2015
RDH approaches[15]. The table summarizes the different [4] Rathika R, S.Kumaresan, “Survey on Reversible Data Hiding
approaches of reversible data hiding and the results of Techniques,” 3rd International Conference on Advanced Computing
different approaches are compared in the below table. The and Communication Systems, jan 2016.
capacity and the PSNR is differs from one method to another. [5] Juhi Gupta, Priya Gupta, “Reversible Data Hiding Technique Using
Histogram Shifting,” Computing for Sustainable Global Development
Fig. 4. demonstrates the PSNR comparison for different (INDIACom), 2015 2nd International Conference on, May 2015,pp.
approaches . The embedding capacity of different methods 2114-2119.
are summarized in the below table and Fig. 5. represents the [6] Manisha G et al. “Reversible Data Hiding Technique and its Type, a
embedding capability comparison for RDH approaches. survey”. IOSR Journal of Computer Engineering (IOSR-JCE), PP 43-
48, 2016.
TABLE III. COMPARITIVE ANALYSIS OF RDH TECHNIQUES [7] Zhicheng Ni et al. “Reversible data hiding. in IEEE Transactions on
Circuits and Systems for Video Technology”, vol. 16, no. 3, pp. 354-
60 362, 2006.
50 [8] Jitha Raj.T et al. “A Survey Paper on Various Reversible Data Hiding
Techniques in Encrypted Images”, IEEE International Advance
40 Computing Conference (IACC), PP 1139-1143, 2015
[9] Rathika R, S.Kumaresan, “Survey on Reversible Data Hiding
30
PSNR

Techniques,” 3rd International Conference on Advanced Computing

and Communication Systems, jan 2016.
20
[10] Juhi Gupta, Priya Gupta, “Reversible Data Hiding Technique Using
10 Histogram Shifting,” Computing for Sustainable Global Development
(INDIACom), 2015 2nd International Conference on, May 2015,pp.
0 2114-2119.
LSB DE PE HS VQ
[11] M. Jeni, Dr. S. Srinivasan, “Reversible Data Hiding in Videos Using
Low Distortion Transform,” Information Communication and
RDH Methods Embedded Systems (ICICES), 2013 International Conference on, Vol
5, issue 5, april 2013.
[12] Dr. E. Kavitha et al., “A Comparative Case Study on The Various
Techniques of Reversible Data Hiding”, Volume VI, Issue IV, PP 81-
84, April 2017
Fig. 4. Comparison of PSNR for Different Approaches [13] Aziz F et al. “Reversible data hiding techniques with high message
embedding capacity in images”. PLoS ONE 15(5): e0231602, 2020.

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

2024 Second International Conference on Advance in Information Technology (ICAIT-2024)

[14] Harshila Gawali et al. “Review of Reversible Data Hiding [20] Asha Jose et al. “Comparative analysis of reversible data hiding
Techniques”. International Research Journal of Engineering and scheme”. IET Image Process, 2020, Vol. 14 Issn 10, pp. 2064-2073,
Technology (IRJET), Volume: 02 Issue: 09, 2015. 2020.
[15] Kamil Khudhair et al. “Secure Reversible Data Hiding Using Block-
Wise Histogram Shifting”. Electronics 12, 1222, 2023.
[16] He B, Chen Y et al. “A novel two-dimensional reversible data hiding
scheme based on high-efficiency histogram shifting for JPEG
images”. International Journal of Distributed Sensor Networks.
2022;18(3).
[17] Osamah M et al. “An Overview of Reversible Data Hiding Schemes
based on Difference Expansion Technique”. First International
Conference on Software Engineering & Computer Systems, PP 741-
746, 2009.
[18] Bini A.A et al. “An Improved Reversible Data Hiding Through
Encryption Scheme with Block Prechecking”, Procedia Computer
Science, Volume 171, Pages 951-958, 2010.
[19] Mohammadi, A et al. “Reversible data hiding in encrypted images
using histogram modification and MSBs integration”. Multimedia
Tools Appl. 2023.

979-8-3503-8386-7/24/$31.00 ©2024 IEEE