Noise Attacks Research Articles

Optical double-image encryption based on Gaussian apertured reality-preserving fractional Mellin transform

A novel optical double-image encryption scheme based on Gaussian apertured reality-preserving fractional Mellin transform (GARPFrMT) is proposed. In terms of optical image encryption, GARPFrMT has several attractive advantages: (1) The GAPRFrMT can be used to scramble the value of two images simultaneously; (2) GARPFrMT is constructed in diffraction domain to ensure that the ciphertext is real; (3) Owing to the nonlinearity, this transform can ensure that the proposed encryption system can be kept from the potential insecurity; and (4) A Gaussian lens is applied to improve intensity distributions of the output light. Experimental results demonstrate that the proposed algorithm is feasible and robust by a lot of attack analysis, i.e. sensitivity attack analysis (>99.60%), differential attack analysis (NPCR >99.6072 and UACI >33.5681%), entropy values (almost 7.99928), statistical attack (correlation coefficient values are almost 0), noise attack, cropping attack and NIST analysis.

Key Independent Image Deciphering using Neighbourhood Similarity Characteristics and Divide-and-conquer Attack

Aim: The aim of the paper is to analyse the security strength of image encryption schemes which are based on pixel rotation and inversion functions. The key independent image decryption methodologies are presented to obtain original images with intelligible contents from encrypted images using neighbourhood similarity characteristics and divide-and-conquer attack. Background: The efficiency and security strength of secure communication of sensitive data depend on the computing resources and cryptographic strength of encryption schemes. An encryption scheme is cryptographically strong if it does not leave any weakness, vulnerability or pattern which could be exploited by cryptanalyst to obtain the original image from an encrypted image. Prior to the use of any image encryption scheme for multimedia security applications, it should be analysed for its security strength to ensure the safety of information so that an adversary could not extract intelligible information from encrypted image data. A number of encryption schemes developed for image security applications and claimed highly secure but some of these are cryptanalysed successfully and found insecure. Objective: The analysis of image ciphers which encrypt plain images by transforming its pixels using circular rotation or inversion function in a random fashion is carried out for decrypting encrypted images to obtain original images. The encryption schemes, namely ‘Chaotic Image Encryption (CIE)’ and ‘Graphical Image Encryption (GIE)’, were reported secure but we find that these schemes are insecure which can be exploited to obtain meaningful information from the ciphered images. We apply the similarity characteristics of images to mount cryptanalytic attacks on these ciphers and obtain original images without any knowledge of the encryption/decryption keys. These encryption schemes encrypting the specified region-of-interest (ROI) are also analysed to decrypt ROI encrypted images. Method: The methodology of decryption is key independent and based on divide-and-conquer strategy to obtain original images from the given encrypted images. It utilizes the neighbourhood similarity of pixels in an image which is measured in terms of pixel-to-pixel difference between adjacent pixels for pixel inversion based image cipher (GIE) and line-to-line correlation between adjacent lines for pixel rotation based image cipher (CIE). The ROI encrypted and masked encrypted images are also decrypted. Results: Experimental test results show that the decrypted images obtained are quite intelligible and one can understand the contents of decrypted images. It is also seen that an image cipher encrypting the ROI can be decrypted by utilizing unencrypted region surrounding encrypted ROI part of an image. Conclusion: It has been shown that CIE, GIE, ROI and masked encryption schemes reported for image security applications are insecure and not providing adequate security. Such encrypted images can be decrypted successfully without any key knowledge with high intelligibility by considering image similarity characteristics of neighbouring pixels and applying divide-and-conquer attack strategy. Future work: The key independent decryption methodology can be considered to cryptanalyse the encryption schemes under noise attack scenario as future work to see the applicability of decryption methods with respect to increase in noise in encrypted images. Moreover, other modern encryption schemes based on pixel inversion and rotation functions can be analysed for their security strength.

RETRACTED ARTICLE: Hybrid Speech Steganography System using SS-RDWT with IPDP-MLE approach

Generally, defence applications need the high secured communication over the channel. Most of the defence-based communications are performed through the audio and speech signals in the secured environment. To provide security to these digital media and speech signals, considerably advanced technologies are implementing in the last two decades. However, digital media transmission over the secured and unsecured channels creating several big problems in defence data transfer applications due to equivalence of works, plasticity, ease of replication and unauthorized use in the digital format of data. The possible solution is to secure digital media by cover writing the speech signals by using speech steganography, so the digital media effectively hid over the speech. To overcome these problems, this article is mainly focussing on securing the speech and message through the hybrid speech steganography approach. This system is implemented with the spread spectrum-based redundant discrete wavelet transform (SS-RDWT) with additional pause removal properties. The input cover speech signal contains a larger size, and system requires the higher power usage, computational time and extra storage capacity. Thus, an intelligent pause detection protocol (IPDP) is implemented to remove the pauses and reduces the size of the speech signal. The performance of the IPDP protocol is improved by the maximum likelihood estimation (MLE) procedure. Finally, SS-RDWT speech steganography is implemented along with the IPDP-MLE-based pause removal mechanism. The simulation results show that the proposed method gives better performance against various noise attacks and improved the robustness, imperceptibility and security as compared to the conventional approaches.

Adaptive Data Augmentation to Achieve Noise Robustness and Overcome Data Deficiency for Deep Learning

Artificial intelligence technologies and robot vision systems are core technologies in smart factories. Currently, there is scholarly interest in automatic data feature extraction in smart factories using deep learning networks. However, sufficient training data are required to train these networks. In addition, barely perceptible noise can affect classification accuracy. Therefore, to increase the amount of training data and achieve robustness against noise attacks, a data augmentation method implemented using the adaptive inverse peak signal-to-noise ratio was developed in this study to consider the influence of the color characteristics of the training images. This method was used to automatically determine the optimal perturbation range of the color perturbation method for generating images using weights based on the characteristics of the training images. The experimental results showed that the proposed method could generate new training images from original images, classify noisy images with greater accuracy, and generally improve the classification accuracy. This demonstrates that the proposed method is effective and robust to noise, even when the training data are deficient.

An image encryption algorithm based on hyperchaotic system and DNA coding

A chaotic image encryption scheme combining block scrambling and DNA coding is proposed. The chaotic sequences generated by Chen hyperchaotic system are used for inter-block index scrambling and intra-block Fisher-Yates scrambling. Then the index sequence is used to further scramble the image to further enhance the encryption effect. The pixel value of the image is initially changed by chaotic sequence, and diffused by different DNA coding rules and different operation rules corresponding to chaotic sequence. After decoding the DNA matrix, the final encrypted image is obtained. The initial values and parameters of the chaotic system are generated by SHA-512 hash function and plaintext image. The experimental results and various security analysis show that the algorithm has good encryption effect and can resist common attacks such as chosen-plaintext attacks, clipping attacks and noise attacks.

Selective image encryption method based on dynamic DNA coding and new chaotic map

In order to improve the various security indicators and encryption and decryption speeds of dynamic DNA coding in the field of image encryption, a selective image encryption method based on new chaotic map and dynamic DNA coding is proposed. In this paper, we first constructed a one-dimensional chaotic map with higher Lyapunov exponent and better dynamic behavior, and then improved the local graph structure algorithm to make it better to select the image region, and use the constructed chaotic map generates pseudo-random sequences to dynamically encode and manipulate the selected area of DNA, and finally generate a complete encrypted image. Experimental results show that compared with the existing methods, the proposed algorithm has a sufficiently large key space, better statistics and differential characteristics, and can resist common attacks such as chosen plaintext attacks, noise attacks, and differential attacks. Therefore, the algorithm has higher security, and significantly improves the encryption and decryption speed of dynamic DNA coding in the field of image encryption, which has a broad application prospect in the field of image encryption.

Probabilistic robustness estimates for feed-forward neural networks

Robustness of deep neural networks is a critical issue in practical applications. In the general case of feed-forward neural networks (including convolutional deep neural network architectures), under random noise attacks, we propose to study the probability that the output of the network deviates from its nominal value by a given threshold. We derive a simple concentration inequality for the propagation of the input uncertainty through the network using the Cramer–Chernoff method and estimates of the local variation of the neural network mapping computed at the training points. We further discuss and exploit the resulting condition on the network to regularize the loss function during training. Finally, we assess the proposed tail probability estimates empirically on various public datasets and show that the observed robustness is very well estimated by the proposed method.

Secure and covert communication using steganography by Wavelet Transform

This paper presents a hybrid image steganography along with cryptography algorithm based on LS (Logical & Statistical) encryption scheme and lift wavelet domain of digital images. Currently, cryptography data can be easily intercepted by an intruder during transmission. So it may be treated as ineffectual in terms of security. In such a scenario, the proposed algorithm provides secured transmission by using cryptography and steganography techniques. In this algorithm a new LS encryption scheme is implemented for converting secret message into ciphertext. It is being done by using several parameters like 'data indicator', 'count indicator', 'gray code' and novel 'data conversion' techniques. The 'gray code' and novel 'data conversion' techniques are used to recover the inherent information at the receiver and to improve the security of the secret data. Moreover, it reduces the ciphertext length compared to the length of original secret data. Hence this scheme improves the payload capacity and helps to improve the quality metrics like Number of Pixels Change in Rate (NPCR), Peak Signal to Noise Ratio (PSNR), and Correlation (CORR) in the stego image. Furthermore, steganography scheme provides an ability to tackle the various attacks like noise, rotational, histogram and visual attacks. It is also shown that how the original message is recovered back from stego image successfully.

Multi-Image Encryption Based on Compressed Sensing and Deep Learning in Optical Gyrator Domain

In this paper, a multi-image encryption scheme based on compressed sensing (CS) and deep learning in the optical gyrator domain is proposed. Firstly, multiple plaintext images are compressed by CS to obtain multiple measurements, and then the pixels of each measurement are scrambled by using a chaotic system. Secondly, the scrambled measurements are combined into a matrix and diffused by XOR operation with a chaotic matrix. Finally, the diffused matrix is encoded with a random phase and an optical gyrator transform to obtain a complex-valued matrix, and the amplitude of the complex-valued matrix is taken as the ciphertext. In decrypt, plaintext images are reconstructed from the CS measurements by a neural network, which achieves high reconstruction speed and quality compared with the traditional algorithm. Especially, the data amount of plaintext images can be compressed by up to 8 times while achieving high decryption quality. To our best knowledge, CS reconstruction algorithms based on deep learning is firstly used for image encryption. Moreover, the proposed scheme is highly robust against occlusion, noise, and chosen-plaintext attack.

Tensor Methods in Computer Vision and Deep Learning

Tensors, or multidimensional arrays, are data structures that can naturally represent visual data of multiple dimensions. Inherently able to efficiently capture structured, latent semantic spaces and high-order interactions, tensors have a long history of applications in a wide span of computer vision problems. With the advent of the deep learning paradigm shift in computer vision, tensors have become even more fundamental. Indeed, essential ingredients in modern deep learning architectures, such as convolutions and attention mechanisms, can readily be considered as tensor mappings. In effect, tensor methods are increasingly finding significant applications in deep learning, including the design of memory and compute efficient network architectures, improving robustness to random noise and adversarial attacks, and aiding the theoretical understanding of deep networks. This article provides an in-depth and practical review of tensors and tensor methods in the context of representation learning and deep learning, with a particular focus on visual data analysis and computer vision applications. Concretely, besides fundamental work in tensor-based visual data analysis methods, we focus on recent developments that have brought on a gradual increase of tensor methods, especially in deep learning architectures, and their implications in computer vision applications. To further enable the newcomer to grasp such concepts quickly, we provide companion Python notebooks, covering key aspects of the paper and implementing them, step-by-step with TensorLy.

Development of payload capacity enhanced robust video watermarking scheme based on symmetry of circle using lifting wavelet transform and SURF

Using simple image and video editing software it is relatively easy to pirate the digital contents at the cost of huge revenue loss to the content owner. Entertainment is the key industry that is overwhelmed by huge financial loss due to digital piracy. WhatsApp, Facebook, Instagram, Pinterest and Twitter are some of the massive social media platforms that are used by millions of users. Many of the users share digital content without giving credit to the owner. Digital watermarking is an efficient technique to protect the digital rights and legal compensation of content owners. Due to the poor selection of watermarking locations, most of the state-of-art methods were fall short to preserve the relationships among the host data. Hence such methods were not efficient to provide safeguards against digital piracy. Therefore, this paper proposes a digital video watermarking technique that analyzes, extracts and processes the useful patterns or features of data based on the symmetry of the circle using 2D-Lifting Wavelet Transform (2D-LWT) and Speeded-Up Robust Features (SURF). SURF method is used to obtain five invariant feature points on an approximation part of 2D-LWT utilizing the luminance component of each video frame. The invariant points serve as centers of the circles used to find the embed locations. The maximum intensity value point in all the points on the circumference of the 1/8 part of the circle is then calculated to serve as an embedding location. Three other points in different quadrants are simply obtained using the concept of symmetry of the circle. After applying temporal desynchronization, compression and insertion of noise attacks the video is assessed through subjective and objective evaluation metrics. The results reveal a high degree of correlation (close to one) and good visual transparency which proves that the proposed scheme is superior for practical applications.

Unequal modulus decomposition and modified Gerchberg Saxton algorithm based asymmetric cryptosystem in Chirp-Z transform domain

In this paper, we have presented an image encryption method in Chirp-Z transform domain using unequal modulus decomposition (UMD) and modified Gerchberg–Saxton (GS) algorithm. The proposed encryption scheme is highly sensitive to the encryption keys and the modified GS algorithm introduces an additional layer of security. The validity of the proposed method is tested with various grayscale and binary images and the numerical simulation results are demonstrated for ‘Cameraman’, ‘Medical’ and Binary ‘CUH’ images. The presented results confirm the robustness of the proposed method against various existing attacks such as, the noise attack, special attack, statistical attack, and brute force attack. A comparative analysis with existing similar methods is also performed and the enhanced security and efficiency of the proposed method is verified.

Non-Linear Hopped Chaos Parameters-Based Image Encryption Algorithm Using Histogram Equalization.

Multimedia wireless communications have rapidly developed over the years. Accordingly, an increasing demand for more secured media transmission is required to protect multimedia contents. Image encryption schemes have been proposed over the years, but the most secure and reliable schemes are those based on chaotic maps, due to the intrinsic features in such kinds of multimedia contents regarding the pixels’ high correlation and data handling capabilities. The novel proposed encryption algorithm introduced in this article is based on a 3D hopping chaotic map instead of fixed chaotic logistic maps. The non-linearity behavior of the proposed algorithm, in terms of both position permutation and value transformation, results in a more secured encryption algorithm due to its non-convergence, non-periodicity, and sensitivity to the applied initial conditions. Several statistical and analytical tests such as entropy, correlation, key sensitivity, key space, peak signal-to-noise ratio, noise attacks, number of pixels changing rate (NPCR), unified average change intensity randomness (UACI), and others tests were applied to measure the strength of the proposed encryption scheme. The obtained results prove that the proposed scheme is very robust against different cryptography attacks compared to similar encryption schemes.

A novel cross cosine map based medical image cryptosystem using dynamic bit-level diffusion

Development in networking technology has made the remote diagnosis and treatment of patients a reality through telemedicine. At the same time, storing and transmitting medical documents over an insecure network has always been a daunting challenge. Literature shows that the existing medical image cryptosystems suffer from serious shortcomings. The present work attempts to address this crucial area and propose a suitable solution. A novel hybrid 2-Dimensional Cross Cosine Map (2D-CCM) with an improved chaotic behaviour cryptosystem is proposed. Chaotic series generated by proposed 2D-CCM is utilized in confusion and diffusion architecture. A new dynamic Bit-Flipping in diffusion approach increases the complexity to achieve good encryption results. Simulation and security level assessment is carried out by executing Statistical and differential attack analysis, key sensitivity and exhaustive attack analysis. Robustness is evidenced by the result of noise and crop attack analysis. In addition, SHA-256 is utilized to feed the seed key value of 2D-CCM to resist plaintext attacks. All the assessment and comparison results illustrate that the cipher possesses enhanced security and efficiency than the existing state of the art. Hence the proposed cipher is absolutely apt for secure medical image communication.

Compressive Sensing Theory for Improving the Robustness and the Security of the Discrete Wavelet Transform-Singular Value Decomposition Watermarking Scheme

Digital watermarking has been suggested as a solution to protect copyright and ownership of digital images. The effective watermarking scheme should satisfy imperceptibility, robustness, capacity and security requirements. In this research, the security issue and the lack of reconstruction quality of the watermark logo of the integration of Discrete Wavelet Transform (DWT), Singular Value Decomposition (SVD) and Compressive Sensing (CS) watermark technique are addressed. Wilkinson measurement matrix is proposed to increase the robustness and improve the reconstruction quality of the watermark logo in the DWT and SVD based on the CS watermarking scheme. Additionally, the security issue of the SVD is solved by sending a scrambled compressed logo then extracting U, V matrices from this compressed logo at the decoder. The proposed method not only increases security but also reduces the size of the transmitted watermark logo. The robustness of the proposed algorithm is evaluated in the presence of different attacks like signal processing, compression, geometrical and noise attacks. The simulation results demonstrate that the proposed technique is more robust than the conventional techniques for all attacks, as well as its robust security. Moreover, the quality of the watermarked image is measured and the results show that the imaging performance of the proposed technique is approximately 0.7 to 4 dB better than that of the conventional methods. Consequently, the proposed watermarking scheme proves its ability to achieve the demands of security, imperceptibility and robustness against different attacks.

A watermarking scheme for source authentication, ownership identification, tamper detection and restoration for color medical images

This paper proposes a blind watermarking scheme that has ability of confidentiality, source authentication, ownership identification, integrity, tampering detection and restoration of the medical images. It embeds the hospital logo, electronic patient record and perceptual hash value of region of interest in the mid-frequency coefficients of discrete cosine transform of the region of non-interest of medical image. The region of interest of the image is restored against attacks including the desynchronization attacks and impulse noise attack, and the severity of tampering due to any attack is measured by comparing the difference between the original perceptual hash value of the region of interest and the extracted perceptual hash value of region of interest of watermarked image. Restoration is done that tampers the region of interest of the images. Number of medical image watermarking schemes that show resilient to many different types of singular and hybrid attacks is very few. The proposed method has inbuilt restoration schemes against different attacks such as rotation, scaling, translation, shearing, horizontal reflection, vertical reflection and impulse noise attacks. Comparison with the latest and state-of-art medical image watermarking schemes shows that the performance of the proposed method is superior to other methods.

Tensor Dropout for Robust Learning

CNNs achieve high levels of performance by leveraging deep, over-parametrized neural architectures, trained on large datasets. However, they exhibit limited generalization abilities outside their training domain and lack robustness to corruptions such as noise and adversarial attacks. To improve robustness and obtain more computationally and memory efficient models, better inductive biases are needed. To provide such inductive biases, tensor layers have been successfully proposed to leverage multi-linear structure through higher-order computations. In this paper, we propose tensor dropout, a randomization technique that can be applied to tensor factorizations, such as those parametrizing tensor layers. In particular, we study tensor regression layers, parametrized by low-rank weight tensors and augmented with our proposed tensor dropout. We empirically show that our approach improves generalization for image classification on ImageNet and CIFAR-100. We also establish state-of-the-art accuracy for phenotypic trait prediction on the largest available dataset of brain MRI (U.K. Biobank), where multi-linear structure is paramount. In all cases, we demonstrate superior performance and significantly improved robustness, both to noisy inputs and to adversarial attacks. We establish the theoretical validity of our approach and the regularizing effect of tensor dropout by demonstrating the link between randomized tensor regression with tensor dropout and deterministic regularized tensor regression.

A new structure-preserving quaternion QR decomposition method for color image blind watermarking

Most of the existing color image watermarking schemes are designed to mark each color channel individually, which ignores the correlation of different color channels and the synchronous embedding of watermarks. In this paper, we propose a robust blind watermarking scheme based on quaternion QR decomposition (QQRD) for color image copyright protection, while using algebraic structure-preserving method to release its computational complexity. In watermark embedding process, QQRD is applied to non-overlapping 4×4 image blocks in a whole manner, whereupon three strong correlation coefficient pairs determined by the Normalized Cross-Correlation (NC) method are modified to embed color watermark within factorized matrices. This avoids layered processing that causes abnormal sensitivity to attacks and less correlation that causes excessive modification of coefficient values. Numerical results show that the newly proposed watermarking scheme has strong ability to resist most image processing attacks and is error-free under no attack, brightness adjustment attack and Poisson noise attack. Its superiority over other state-of-the-art methods is also experimentally demonstrated.

A secure video steganography scheme using DWT based on object tracking

ABSTRACT Videos are nowadays the most frequent and easy mode of communication over the internet; the reason behind the growth is the accessibility to video processing software available on the internet. Video steganography is a field where the data is embedded in video keeping the visual quality of the video intact. This study presents a new video steganography scheme with a stable trade-off between robustness and imperceptibility using 2D-DWT (Discrete Wavelet Transform) based on object detection and tracking. The main contribution of this paper includes embedding of secret data in the moving objects after applying object detection for the video frames where the secret bits are embedded in middle frequency sub-bands after applying 2D-DWT. To highlight the effectiveness of the proposed scheme, experimental results are carried out both quantitatively and qualitatively where quantitative analysis is done using different metrics such as PSNR, SSIM, BER and qualitative analysis is done using visual results of the frames. The experimental results illustrated that the proposed approach outperforms existing techniques in terms of qualitative and quantitative evaluation with high imperceptibility and robustness against noise attack. Eventually, the scheme has also been tested against existing steganalysis techniques to ensure the security of the proposed scheme.

Medical Image Protection Algorithm Based on Deoxyribonucleic Acid Chain of Dynamic Length

Current image encryption algorithms have various deficiencies in effectively protecting medical images with large storage capacity and high pixel correlation. This article proposed a new image protection algorithm based on the deoxyribonucleic acid chain of dynamic length, which achieved image encryption by DNA dynamic coding, generation of DNA dynamic chain, and dynamic operation of row chain and column chain. First, the original image is encoded dynamically according to the binary bit from a pixel, and the DNA sequence matrix is scrambled. Second, DNA sequence matrices are dynamically segmented into DNA chains of different lengths. After that, row and column deletion operation and transposition operation of DNA dynamic chain are carried out, respectively, which made DNA chain matrix double shuffle. Finally, the encrypted image is got after recombining DNA chains of different lengths. The proposed algorithm was tested on a list of medical images. Results showed that the proposed algorithm showed excellent security performance, and it is immune to noise attack, occlusion attack, and all common cryptographic attacks.