Page 50 - 2024-Vol20-Issue2
P. 50
46 | Alobaidi & Mikhael
tal images. Common methods include Least Significant Bit texture images, and 9 aerial images. These images were
(LSB) insertion [5], where data bits (the information to be hid- tested with different dimensions, in particular 2d × 2d pixels
den) are stored in the least significant bit(s) of the image pixels where d = [7, 8, 9, and 10]. The secret message size varied
(called the cover image that is the image in which you want to between 2KB and 16KB. Evaluation metrics included PSNR,
hide the secret information), and the use of spread spectrum MSE, Structural Similarity Index (SSIM), and Normalized
techniques to distribute the hidden information across the im- Cross-Correlation (NCC). The results demonstrated that this
age to obtain the encoded image, or stegoimage. In Audio approach outperforms existing techniques. Further details
Steganography, hidden information is concealed within audio regarding the results can be found in Section IV. in this work.
files. Techniques like phase coding, echo hiding, and audio In [13], an examination of existing approaches, analyze cur-
masking exploit the characteristics of sound to embed secret rent trends, and address the obstacles encountered in related
messages. In Text Steganography technique, information is research. It also included an examination of publicly acces-
hidden within textual content. Methods include employing sible databases commonly employed in these studies and the
invisible ink, modifying font styles, or utilizing hidden spaces evaluation measures utilized. Furthermore, the paper pre-
or punctuation marks to encode data. sented a comparative analysis of the performance of different
On the other hand and as steganography techniques continue methods and engages in a discussion regarding the identified
to evolve, so do steganalysis [6] methods used to detect and gaps, advantages, and disadvantages of the approaches utilized
analyze hidden messages. Steganalysis involves the applica- in the present research. In [14], an approach is introduced that
tion of statistical analysis [7], machine learning algorithms, utilizes k least significant bits (LSB) coding to hide an image.
and forensic techniques to identify the presence of stegano- This k-LSB-based method employs a specific number of least
graphic content. Some common steganalysis methods include significant bits to conceal the image. To decode the hidden
Statistical Analysis in which Steganalysis algorithms analyze image, a region detection operation is performed to identify
statistical properties of carrier files to detect deviations from the blocks that contain the concealed image.
the expected patterns. These include examining pixel inten- The resolution of the resulting stego image may be impacted,
sity distributions, correlation between neighboring pixels, and so an image quality enhancement technique is employed to
frequency domain analysis. improve the resolution. In order to showcase the effective-
In addition to the first technique, Machine Learning-Based ness of this proposed approach, a comparison is made against
Steganalysis in which algorithms are trained to distinguish be- several state-of-the-art methods. In [15], a robust and se-
tween normal and steganographic content by learning patterns cure video steganographic algorithm was proposed in the
from a large database of known carriers. Support Vector Ma- discrete wavelet transform (DWT) and discrete cosine trans-
chines (SVMs) [8], Artificial Neural Networks (ANNs) [9], form (DCT) domains, based on the multiple object tracking
and Random Forests [10] are commonly used in machine (MOT) algorithm and error correcting codes. The secret mes-
learning-based steganalysis. Finally, Visual Inspection where, sage was preprocessed by applying both Hamming and Bose,
in some cases, visual inspection by trained experts is em- Chaudhuri, and Hocquenghem codes to encode the secret data.
ployed to identify visual anomalies or irregularities in the Initially, the motion-based MOT algorithm was implemented
carrier files that may indicate the presence of hidden informa- on host videos to differentiate the regions of interest in the
tion. moving objects. Subsequently, the data hiding process was
MSE (Mean Squared Error), RMSE (Root Mean Squared Er- performed by hiding the secret message within the DWT and
ror), SSIM (Structural Similarity Index), PSNR (Peak Signal- DCT coefficients of all motion regions in the video, based
to-Noise Ratio) are commonly used metrics in image and on foreground masks. The experimental results demonstrated
video processing to evaluate the quality or fidelity of a re- that the suggested algorithm not only improved the embedding
constructed or compressed signal compared to the original capacity and imperceptibility, but also enhanced its security
signal [11]. and robustness by encoding the secret message and withstand-
In [12], introduced a novel steganography method employing ing various attacks.
LSB. The paper also provided information about recent rele- In [16], a work presented a novel technique for image steganog-
vant approaches. The process involved flipping and transform- raphy based on Huffman Encoding. Two 8-bit gray level im-
ing the image and then dividing it into its 3 color channels: ages of size M X N and P X Q were used as the cover image
red, green, and blue. The blue channel is rearranged using and secret image respectively. Huffman Encoding was per-
the Magic Matrix, a built-in MATLAB function, to hide a formed over the secret image/message before embedding, and
secret message. To enhance security, Multi-Level Encryp- each bit of the Huffman code of the secret image/message was
tion (MLEA) algorithm is utilized. The proposed technique embedded inside the cover image by altering the least signif-
is evaluated using 12 color images, 9 grayscale images, 9 icant bit (LSB) of each pixel’s intensity in the cover image.