Enhanced Steganographic Framework for Secure Communications

The term "steganography" originates from the Greek word "steganos," which translates to "hidden writing." The term "steganos" consists of two constituent elements, namely "secret" and "graphic," with the latter denoting the act of writing. In contrast, Steganography refers to the practice of covertly embedding confidential messages or textual content within various types of media, such as textual documents, photographs, videos, or musical compositions. There exists a significant degree of overlap among the concepts of "Watermarking", "Steganography", and "Cryptography" [1, 2]. Watermarking serves to authenticate the integrity of the communication, whereas steganography conceals it, and cryptography obfuscates it. According to the study conducted by Sasmal and Mula [1], in the field of steganography, it is imperative for the sender to carefully select a suitable message carrier prior to commencing the concealing process. In order to ensure the secure encryption of authentic data, it is imperative to carefully choose a reliable steganographic approach [2, 3]. The concealed message may, after that, be transmitted to the intended recipient through several contemporary communication modalities by the sender. Upon receiving the message, the recipient is required to employ the appropriate extraction procedure in order to decipher the concealed contents. Different steganographic approaches are employed to provide security, depending on the specific type of carrier being utilized. According to the study conducted by Ali and Uddin in 2019, the primary purpose of steganography is to mitigate the risk of attracting attention to the process of transmitting confidential information. However, the efficacy of the technique of concealing information becomes insufficient when an observer is capable of detecting any alterations in the transmitted data [1]. According to the source provided, Steganography and Cryptography exhibit notable similarities since they have the common objective of safeguarding sensitive information through comparable means. Based on the latest definition, Steganography refers to the act of covertly embedding data or documents within various forms of digital media, such as photographs, audio files, movies, and network transmissions. The prevailing method in steganography is the utilization of the “Least Significant Bit (LSB)” substitution technique. The fundamental concept of LSB replacement involves the concealment of confidential information among bits that possess the least significance, ensuring that the original pixel value remains unaltered [3]. Steganography techniques can be categorized into various types based on the type of cover media employed for the purpose of concealing confidential information. These types include text steganography, image steganography, audio steganography, network protocol steganography, and video steganography [3, 4].

Unlike prior LSB-only approaches, this framework integrates multi-stage encryption and dynamic red, green, and blue (RGB) embedding, significantly improving security.

In these times of rapid technology development, besides the increase in software application complexity, the availability, integrity, and security of data have been compromised. Thus, measures to protect these systems and data need to be introduced. Arya and Soni presented an improved approach to the LSB substitution method to merge the hidden data of an image onto the cover image in such a way that the embedded information does not draw the attention of the observer. This is achieved by replacing the LSB of each pixel in the original image. They also tested the two techniques previously stated by the use of pictures of different lengths and widths sizes of other file formats (bmp, jpg, png). They computed the Peak Signal-to-Noise Ratio (PSNR) and Mean Squared Error (MSE) for those techniques to examine their information-hiding capabilities [8].

Gouthamanaath and Kangaiammal [9] proposed a new approach in binary image steganography. In this steganography technique, the invariant texture patterns are created from a binary image, which can be rotation, complementation, and mirroring, etc. A computation was given using the suggested measurement mentioned above. The proposed stenographic approach is sufficiently statistically strong, of high visual quality, and, thus, easy to watermark [9]. In another attempt, recent work by Thanki and Borra [10] in 2018 presented another type of steganography approach based on “Finite Ridgelet Transform (FRT)”, “Discrete Wavelet Transform DWT”, and “Arnold scrambling”. For the purpose of safety assurance, the “FRT” was selected for color image protection. The proposed technique attempts to embed the secret color image within the inserted cover color image by using Arnold scrambling and, hence, obtains the stego color image.

In the 2019 research study [11], several metrics were used to measure the similarity between the stego image (SI) and the carrier image. These metrics included “Mean Squared Error (MSE)”, “Peak Signal-To-Noise Ratio (PSNR)”, “Structural Similarity Index (SSIM)”, histograms, CPU time, and “Feature Similarity Index Measure (FSIM)”. Their study and experimentation showed that their proposed strategy was better in speed and efficiency compared to the conventional LSB techniques. Astuti et al. [12] introduced the use of the bit interchanging technique for the evaluation of RGB features of the image format. Where per pixel is a message capacity of 1 bit. The introduction of more layers into the color image brings more diversity in the result for the imperceptibility test. The bit-flipping imperceptibility technique has also been tested effectively on color images [12].

Rahman et al. [13] had gone one step closer in this direction, suggesting a new steganography method inside the RGB color model that would have more robust security features compared to earlier technology. Various measures serve to evaluate the quality of the image. The results aim to have stronger robustness, imperceptibility, and security than prior techniques, thus validating the efficiency of the experimental mission. In terms of PSNR correlation, an overall average value of 3.6701 percent higher was obtained with the proposed method [13]. In the study of Kaur et al. [14], the method of steganography is a means of digital image watermarking. This project aims to safeguard the image file during the transmission process from undesirable intruders. This paper proposes a novel method known as Image Hiding Encryption and Decryption (IHED) to convert an image into an encrypted form and then decode it. Also, a model called the “Mid Search African Buffalo Model (MSABM)” is exploited to determine the “Mid-Frequency MF” values prior to the process of encoding [14].

For example, the research of Sahu and Swain [15] recently advanced work in the domain of LSB image steganography to enhance the robustness and steganographic performance of integration capacity to keep secret information [15]. For example, the work of Almazaydeh [16] embedded RGB images in steganography to accumulate more security during data transfer over the internet. The cover image is a 24-bit RGB image with embedded information. The X-Box mapping consists of 16 possible values having multiple bins. "X" is a variable that ranges from 0 to 9. The X-boxes are derived from the LSBs of the binary image. The technique of mapping is a method to enhance imperceptibility, but the mapping gives challenges too, in the process of extracting information. It is then installed with high-quality digital resistance protocols. The strength of the steganographic image can be measured using the PSNR [16].

As provided in the research of Abuzanouneh and Hadwan [17], a feature selection technique, along with pixel selection methodology, was combined so that the secret signal could be buried. The hidden file is fragmented and embedded in random positions of the stego image to bring more challenges into the steganalysis operation. The secret file contains encrypted units in the form of a binary string. At the same time, the MPPST algorithm generates a complex key based on the key parameters, and this key can identify the units' location within the binary string. To compare endpoint security effectiveness using a newly proposed method to one of the most famous approaches using the LSB technique, MPPST is applied [17].

The classification of the various steganography systems proposed in the study of Dhawan and Gupta [18] is based on mathematical and non-mathematical grounds, and they are designed in their corresponding domains. The stego images are evaluated primarily based on the following considerations: “Image Fidelity (IF)”, payload capacity, MSE, resilience, “Structural Similarity Index (SSIM)”, and “Normalized Cross-Correlation (NCC)”. These properties are of highly significant importance to steganography and must be taken into consideration in the process of application. The paper aimed to compare and evaluate various algorithms based on steganography according to PSNR, MSE, and robustness [18].

In a recent paper, Tang et al. [19] proposed a new method for color image steganography with FIS in a structural field-iterative framework. The fuzzy inference system results, together with the human eye's perceptual sensitivity to the red (R), green (G), and blue (B) color components, allow the data to be dynamically hidden by LSB alternation. The whole process used in bits to embed in the original image includes the particular color plane used in coding and the results of fuzzy reasoning, according to reference [19].

Almawgani et al. [20] proposed a novel article of the well-known steganographic technique based on the “Haar Discrete Wavelet Transform (HDWT)”, “Lempel Ziv Welch (LZW)” algorithm, “Genetic Algorithm (GA)”, and “Optimal Pixel Adjustment Process (OPAP)”. The considered approach divides the original image into non-overlapping blocks of (n x n) pixels. The use of HDWT makes the steganographic image resistant to various attacks. The LZW technique is applied to the secret message to increase the hidden image’s robustness and capacity. The coefficients of the original image in the secret message are encrypted and further compressed using a GA. The “OPAP” decreases the error rate [20].

This work describes the design of a deep neural network (DNN)-based approach to secretly holding ample confidential information. This method will use DNNs from the information deception stage to the data extraction stage. High-resolution color images can secretly be embedded into low-resolution color images. Using the DNNs in the information deception stage to the data extraction stage is known as our approach. Adversarial training, besides being used simultaneously in the system, has also been used to achieve the two networks. The fixed information ratio of the carrier picture and the hidden image is 1:4. The User has given a numeric reference in the study [21].

The model under consideration was executed within the Python software environment. The tests were conducted using a system equipped with an Intel (R) Core (TM) i5 CPU and 8 GB of RAM. The standard color images commonly employed in experimental studies include Lena, Peppers, and Baboon. Experiments used standard 512 × 512 images (Lena, Peppers, Baboon) from the USC-SIPI database, selected for their prevalence in steganography literature. Figures 5-7 depict a conventional cover color image alongside stego images, together with their respective histograms. The figures illustrate that there is a lack of substantial alteration in both the cover and stego pictures, as well as their corresponding histograms. This observation serves as evidence supporting the efficacy of the proposed strategy. Table 1 presents the quality assessment through the values of PSNR and MSE, which are deemed to be within acceptable ranges. Table 2 shows a comparative analysis with state-of-the-art; the proposed method achieves higher PSNR/lower MSE due to optimized LSB embedding and XOR encryption, which minimizes pixel distortion compared to other methods. Additionally, Table 3 and Figure 8 display security analysis and attack resistance. The 40-100% improvement in attack resistance (Table 3) stems from multi-layer encryption disrupting statistical patterns, unlike basic LSB.

The proposed steganographic framework deploys multi-layered security for effective covert communication. It uses the Caesar Cipher for initial encryption, with low computational overhead and real-time applicability, by using a shift value of 3 to scramble the secret message. The encoded message is XORed with a secret key and then embedded in the LSBs of the cover image pixels. This approach of embedding LSB is capable of providing the visual quality of the image and concealing the data efficiently. The extraction and decryption process, with the help of LSB extraction, XOR decryption, and Caesar Cipher decryption, recovers and restores the original message. Although the framework is quite sound and efficient, its use of the Caesar Cipher exposes it to weaknesses against more advanced attacks; therefore, the future incorporation of more robust encryption algorithms, e.g., AES. The following are its critical uses or advantages: Secure data transmission in applications that rely with little degree of importance, for example, secure data transmission in such areas as confidential communication, military operations, and digital rights management, without compromising performance.

Figure 5. The input and output photos of Lena, together with their corresponding histograms

Figure 6. The input and output photos of peppers, together with their corresponding histograms

Figure 7. The input and output photos of Baboon, together with their corresponding histograms

Figure 8. Security analysis and attack resistance

Table 1. MSE and PSNR values for the stenographic images

Table 2. Comparative analysis with state-of-the-art

Table 3. Security analysis and attack resistance

The proposed steganographic framework is, therefore, quite effective in attaining high image quality and securely embedding the secret messages. MSE and PSNR values obtained from the analysis of different images were 0.33 to 0.42 from MSE and 71.02 dB to 72.63 dB from PSNR, respectively. This is minimum visual distortion and high-fidelity level with the original images. The highest quality retention post-embedding was displayed by the image of Peppers, as seen from the lowest MSE and highest PSNR values. The results vindicate the vigorous technique of hiding information in the LSBs used by this framework. It is this practical attribute of the framework in producing stego images, with virtually seen differences, that comes to the fore in providing secure transmission of information, for instance, confidential communications. The results underline the appropriateness of the framework for applications implying covert communication and data security, and hence it is ideal for military communications requiring low latency and medical data transmission, where metadata integrity is critical. The results promote, therefore, the potential of the obtained framework for different security and privacy scenarios.