Separable Data Research Articles

Separable High Capacity Reversible Data Hiding Algorithm for Encrypted Images

This paper presents a separable Reversible Data Hiding Algorithm for Encrypted Images (RDHEI) that consists of three phases. The encryption phase in the algorithm circularly shifts the columns in the image by a random amount, blocks the image into equal and regular blocks and maps them to irregular blocks generated based on Hilbert filling curve, and finally complements a random subset of the blocks. The embedding phase is essentially an adapted version of the modification of the prediction errors algorithm that is applied to each block in the encrypted image independently. In the decryption phase, and since the algorithm is separable, the user can extract the data only, decrypt the image only, or can perform both actions depending on the type of keys he has. When compared to a very similar and recent algorithm, performance evaluation proved the ability of the proposed algorithm in increasing the embedding capacity with reasonable quality of the directly decrypted image. In terms of the security, the analytical and quantitative assessment showed the superiority of the proposed algorithm in protecting the encrypted image.

Adaptive Particle Swarm Optimization Data Hiding for High Security Secret Image Sharing

The main aim of this work is to improve the security of data hiding for secret image sharing. The privacy and security of digital information have become a primary concern nowadays due to the enormous usage of digital technology. The security and the privacy of users’ images are ensured through reversible data hiding techniques. The efficiency of the existing data hiding techniques did not provide optimum performance with multiple end nodes. These issues are solved by using Separable Data Hiding and Adaptive Particle Swarm Optimization (SDHAPSO) algorithm to attain optimal performance. Image encryption, data embedding, data extraction/image recovery are the main phases of the proposed approach. DFT is generally used to extract the transform coefficient matrix from the original image. DFT coefficients are in float format, which assists in transforming the image to integral format using the round function. After obtaining the encrypted image by data-hider, additional data embedding is formulated into high-frequency coefficients. The proposed SDHAPSO is mainly utilized for performance improvement through optimal pixel location selection within the image for secret bits concealment. In addition, the secret data embedding capacity enhancement is focused on image visual quality maintenance. Hence, it is observed from the simulation results that the proposed SDHAPSO technique offers high-level security outcomes with respect to higher PSNR, security level, lesser MSE and higher correlation than existing techniques. Hence, enhanced sensitive information protection is attained, which improves the overall system performance.

Efficient separable reversible data hiding algorithm for compressed 3D mesh models

A typical problem examined in most of the earlier reversible data hiding methods is the vast amount of information to be transmitted towards the recipient end, which significantly limits the applicability of these methods. Hence, the need of data compression is required before transmission. An efficient reversible data hiding algorithm for 3D mesh models in the compressed domain is proposed to ensure secret message privacy and security. The main contribution of proposed algorithm is to hide secret message in 3D mesh model in such a way that mesh vertices and its corresponding topology remained as it is while allowing the receiver to correctly extract secret message and perfectly reconstruct the cover 3D mesh model in a reversible manner. For traversing 3D mesh vertices, a novel mesh traversal algorithm is proposed where the traversal order of stego-mesh is not affected by vertex reordering attack because it is based on the computation of the shortest distance between neighbor vertices regardless of their reference indexes. Experimental study reveals that the proposed method has more embedding rate with comparing image quality than the existing methods. The average PSNR value is 96.40 dB for an embedding rate of 6.94 bits per vertex(bpv) for all test mesh models, demonstrating that the method can give good quality stego mesh models even at high payload also.

Reversal of pixel rotation: A reversible data hiding system towards cybersecurity in encrypted images

Due to privacy and security concerns, the researches of reversible data hiding in encrypted images (RDHEI) have become increasingly important. Conventional schemes vacate the spare room after image encryption (VRAE) suffer from the low embedding rate, high error rate of data extraction, and imperfect image recovery. To address these issues, we propose a separable reversible data hiding scheme for encrypted images that utilizes a novel pixel rotation technique to embed data into fully encrypted images. The block complexities of four decrypted rotation states are considered when recovering image. To realize perfect image recovery, we further devise a lossless version (LPR-RDHEI). Experimental results demonstrate that the proposed PR-RDHEI scheme achieves an embedding rate of 0.4994 bpp on average and ensures lossless data extraction. Meanwhile, the proposed LPR-RDHEI scheme still has a 0.4494 bpp embedding rate on average. The embedding rates of our two schemes are significantly improved compared with state-of-the-arts.

An analytic theory of shallow networks dynamics for hinge loss classification* *This article is an updated version of: Pellegrini F and Biroli G 2020 An analytic theory of shallow networks dynamics for hinge loss classification Advances in Neural Information Processing Systems vol 33 ed H Larochelle, M Ranzato, R Hadsell, M F Balcan and H Lin (New York: Curran Associates)

Neural networks have been shown to perform incredibly well in classification tasks over structured high-dimensional datasets. However, the learning dynamics of such networks is still poorly understood. In this paper we study in detail the training dynamics of a simple type of neural network: a single hidden layer trained to perform a classification task. We show that in a suitable mean-field limit this case maps to a single-node learning problem with a time-dependent dataset determined self-consistently from the average nodes population. We specialize our theory to the prototypical case of a linearly separable data and a linear hinge loss, for which the dynamics can be explicitly solved in the infinite dataset limit. This allows us to address in a simple setting several phenomena appearing in modern networks such as slowing down of training dynamics, crossover between rich and lazy learning, and overfitting. Finally, we assess the limitations of mean-field theory by studying the case of large but finite number of nodes and of training samples.

Automated classification of tursiops aduncus whistles based on a depth-wise separable convolutional neural network and data augmentation.

Whistle classification plays an essential role in studying the habitat and social behaviours of cetaceans. We obtained six categories of sweep whistles of two Tursiops aduncus individual signals using the passive acoustic mornitoring technique over a period of eight months in the Xiamen area. First, we propose a depthwise separable convolutional neural network for whistle classification. The proposed model adopts the depthwise convolution combined with the followed point-by-point convolution instead of the conventional convolution. As a result, it brings a better classification performance in sample sets with relatively independent features between different channels. Meanwhile, it leads to less computational complexity and fewer model parameters. Second, in order to solve the problem of an imbalance in the number of samples under each whistle category, we propose a random series method with five audio augmentation algorithms. The generalization ability of the trained model was improved by using an opening probability for each algorithm and the random selection of each augmentation factor within specific ranges. Finally, we explore the effect of the proposed augmentation method on the performance of our proposed architecture and find that it enhances the accuracy up to 98.53% for the classification of Tursiops aduncus whistles.

Separable reversible data hiding in homomorphic encrypted domain using POB number system

Separable reversible data hiding in homomorphic encrypted domain using POB number system

Separable Reversible Data Hiding Based on Integer Mapping and MSB Prediction for Encrypted 3D Mesh Models

Separable Reversible Data Hiding Based on Integer Mapping and MSB Prediction for Encrypted 3D Mesh Models

Refined PSO Clustering for Not Well-Separated Data

ABSTRACT Cluster analysis is the most popular and often the foremost task in big data analytics as it helps in unearthing hidden patterns and trends in data. Traditional single-objective clustering techniques often suffer from accuracy fluctuations especially when applied over data groups of varying densities and imbalanced distribution as well as in the presence of outliers. This paper presents a multi-phase clustering solution that achieves good accuracy measures even in the case of noisy and not- well-separated data (linearly not separable data). The proposed design combines a two-stage Particle Swarm Optimisation (PSO) clustering with K-means logic and a state-of-the-art outlier removal technique. The use of two different optimisation criteria in the two stages of PSO clustering equips the model with the ability to escape local minima traps in the process of convergence. Extensive experiments featuring a wide variety of data have been carried out and the system could achieve accuracy levels as high as 99.9% and an average of 87.4% on notwell-separated data. The model has also been proved to be robust on eight out of the ten datasets of the Fundamental Clustering Problem Suit (FCPS), a benchmark for clustering algorithms.

An enhanced separable reversible and secure patient data hiding algorithm for telemedicine applications

In telemedicine industry, a standout among the most significant issues is the exchange of Electronic Patient Information (EPI) between patient and a doctor that are remotely connected. A minute change to EPI may result in wrong diagnosis to the patient. With the aim to ensure secure and safe communications in telemedicine framework, an enhanced separable reversible data hiding technique in encrypted domain has been presented here that gives high embedding rate by embedding k, (k≥1) binary bits of patient data in base2k numeral framework using Average Pixel Repetition (APR) method. APR method converts original image of size M×N into encrypted scaled-up image of size (2×M−1)×(2×N−1) such that non-seed pixels are used for data embedding where as seed pixels remain as it is to facilitate reversibility. The experimental study shown that the proposed method gave extremely high embedding rates on both natural and medical images. For all test images, the proposed methodology altogether beated all the compared methodologies in its ability to embed patient information and precisely recover it with maintaining the visual quality of stego images too.

Mixture factor analysis with distance metric constraint for dimensionality reduction

Dimensionality reduction (DR) is a key preprocessing stage in high-dimensional data classification. Traditional linear DR algorithms, e.g., Linear Discriminant Analysis, transform the original data into a low-dimensional subspace with a linear transformation matrix. However, these methods cannot handle complex nonlinearly separable data. Although some nonlinear DR methods, e.g., Locally Linear Embedding, are proposed to solve this problem, most of them are unsupervised, which only focus on the data structure hidden in the original high-dimensional space, rather than maximizing the inter-class separability of the transformed data, thus reducing the classification accuracy. To tackle this challenge, a novel supervised nonlinear DR algorithm, distance metric restricted mixture factor analysis (DMR-MFA), is proposed for high-dimensional data classification. In DMR-MFA, the original data is divided into several clusters, and the generation of original data in each cluster is described via a factor analysis model. Meanwhile, the distance metric constraint (DMC) is used for maximizing the separability of transformed low-dimensional data from different classes. Moreover, the optimal model parameters are learned via the joint optimization of log-likelihood function and DMC loss function, which makes the DMR-MFA possible to obtain the more separable low-dimensional embeddings while accurately describing the original data. Experimental results on synthetic data, benchmark datasets and high-resolution range profile data demonstrate that our method can handle nonlinearly separable data and improves the classification accuracy of data with high dimensionality.

Separable Reversible Data Hiding in Encryption Image with Two-Tuples Coding

Separable Reversible Data Hiding in Encryption Image (RDH-EI) has become widely used in clinical and military applications, social cloud and security surveillance in recent years, contributing significantly to preserving the privacy of digital images. Aiming to address the shortcomings of recent works that directed to achieve high embedding rate by compensating image quality, security, reversible and separable properties, we propose a two-tuples coding method by considering the intrinsic adjacent pixels characteristics of the carrier image, which have a high redundancy between high-order bits. Subsequently, we construct RDH-EI scheme by using high-order bits compression, low-order bits combination, vacancy filling, data embedding and pixel diffusion. Unlike the conventional RDH-EI practices, which have suffered from the deterioration of the original image while embedding additional data, the content owner in our scheme generates the embeddable space in advance, thus lessening the risk of image destruction on the data hider side. The experimental results indicate the effectiveness of our scheme. A ratio of 28.91% effectively compressed the carrier images, and the embedding rate increased to 1.753 bpp with a higher image quality, measured in the PSNR of 45.76 dB.

Surgical Treatment of Osteochondral Lesions of the Tibial Plafond: A Systematic Review and Meta-Analysis.

The literature on osteochondral lesions of the tibial plafond (OLTPs) is sparse. The aim of this study was therefore to provide an overview of clinical and radiological outcomes following treatment of OLTPs. We performed a systematic search of the MEDLINE, Embase, and Cochrane library databases. The review was performed in accordance with the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guidelines and included all original articles on treatment outcomes for OLTPs. The methodological quality of the articles was assessed using the Methodological Index for Non-Randomized Studies (MINORS). Baseline patient and lesion characteristics were pooled and weighted according to the number of lesions per study. The primary outcome was any clinical or patient-reported outcome measure pooled by treatment method when separable data were available. Secondary outcomes were complications, reoperation rates, radiological outcomes, and sport outcomes. The search yielded 2,079 articles, of which 10 studies (1 prospective case series, 1 retrospective comparative study, and 8 retrospective case series) with a total of 175 patients were included. The overall methodological quality of the studies was low. All patients were treated surgically; 96% of the lesions were primary cases (i.e., first-time surgery) and 58% were solitary tibial lesions (i.e., no opposing talar lesion). Arthroscopic bone marrow stimulation was the most frequently used treatment strategy (51%), followed by cartilage transplantation (17%), chondrogenesis-inducing techniques (11%), osteochondral transplantation (3%), retrograde drilling (3%), and mixed (i.e., inseparable) treatments (15%). The clinical outcomes of the different surgical therapies were considered to be moderate to good. The pooled postoperative AOFAS (American Orthopaedic Foot & Ankle Society) score for bone marrow stimulation and osteochondral transplantation was 54.8 (95% confidence interval [CI], 49.5 to 85.0) (n = 14) and 85.3 (95% CI, 56 to 100) (n = 3), respectively. Overall, complications and reoperations were rarely reported. The pooled complication and reoperation rates could only be calculated for bone marrow stimulation and were 5% and 7%, respectively. Surgical interventions for OLTPs appear to yield moderate to good clinical outcomes. Bone marrow stimulation resulted in a moderate AOFAS score. Complications and reintervention rates were found to be low. The current evidence in the literature is limited because of the underreporting of clinical, radiological, and sport data and the heterogenous outcome scores reported. Therapeutic Level IV. See Instructions for Authors for a complete description of levels of evidence.

Free-breathing Accelerated Cardiac MRI Using Deep Learning: Validation in Children and Young Adults.

Background Obtaining ventricular volumetry and mass is key to most cardiac MRI but challenged by long multibreath-hold acquisitions. Purpose To assess the image quality and performance of a highly accelerated, free-breathing, two-dimensional cine cardiac MRI sequence incorporating deep learning (DL) reconstruction compared with reference standard balanced steady-state free precession (bSSFP). Materials and Methods A DL algorithm was developed to reconstruct custom 12-fold accelerated bSSFP cardiac MRI cine images from coil sensitivity maps using 15 iterations of separable three-dimensional convolutions and data consistency steps. The model was trained, validated, and internally tested in 10, two, and 10 adult human volunteers, respectively, based on vendor partner-supplied fully sampled bSSFP acquisitions. For prospective external clinical validation, consecutive children and young adults undergoing cardiac MRI from September through December 2019 at a single children's hospital underwent both conventional and highly accelerated short-axis bSSFP cine acquisitions in one MRI examination. Two radiologists scored overall and volumetric three-dimensional mesh image quality of all short-axis stacks on a five-point Likert scale and manually segmented endocardial and epicardial contours. Scan times and image quality were compared using the Wilcoxon rank sum test. Measurement agreement was assessed with intraclass correlation coefficient and Bland-Altman analysis. Results Fifty participants (mean age, 16 years ± 4 [standard deviation]; range, 5-30 years; 29 men) were evaluated. The mean prescribed acquisition times of accelerated scans (non-breath-held) and bSSFP (excluding breath-hold time) were 0.9 minute ± 0.3 versus 3.0 minutes ± 1.9 (P < .001). Overall and three-dimensional mesh image quality scores were, respectively, 3.8 ± 0.6 versus 4.3 ± 0.6 (P < .001) and 4.0 ± 1.0 versus 4.4 ± 0.8 (P < .001). Raters had strong agreement between all bSSFP and DL measurements, with intraclass correlation coefficients of 0.76 to 0.97, near-zero mean differences, and narrow limits of agreement. Conclusion With slightly lower image quality yet much faster speed, deep learning reconstruction may allow substantially shorter acquisition times of cardiac MRI compared with conventional balanced steady-state free precession MRI performed for ventricular volumetry. © RSNA, 2021 Online supplemental material is available for this article.

Performance Evaluation of Classifiers Created using Elman Back-Propagation and Cascade Feed-forward Neural Networks

Huge data being captured in our day-to-day activities are mostly imbalance. Such data therefore, calls for fast, accurate and robust techniques, through which they could be analyzed in order to fast-track early decision making. A Cascade Feed-forward Neural Networks and Elman Backpropagation are known techniques in neural network domain and their efficacies is therefore tested on separable data in this study. The objective of this study is to evaluate the performance of these techniques in solving a linear classification problem. The linear classification of data involves, splitting of separable data into two distinct clusters. In order to achieve the goal of this study, linear classifiers were created using the two aforementioned techniques. Both network structures were exposed to the same dataset and similar parameter configurations were set for each technique. The model that emanates through each technique was simulated using the set of untrained data. In order to determine the accuracy of each model created through each technique, their Mean Absolute Errors (MAE) was computed. The performance of each model was determined based on the value of MAE. The error computation for the simulated output reveals that, cascade neural network gives an error of 0.0928; while the model created using Elman Backpropagation network gives a relatively lower error of 0.0661. It can be inferred from this study, that, both techniques are capable of fitting accurate classifiers from dataset and specifically, both techniques are very suitable for binary classification of separable data.

Hiding patient information in medical images: an enhanced dual image separable reversible data hiding algorithm for E-healthcare

Hiding patient information in medical images: an enhanced dual image separable reversible data hiding algorithm for E-healthcare

An improved separable reversible patient data hiding algorithm for E-healthcare

In the telemedicine industry, a standout among the most significant issue is the exchange of Electronic Patient Information (EPI) between patient and a doctor that are remotely connected. A minute change to EPI may result in a wrong diagnosis for the patient. To ensure secure and safe communication for telemedicine applications, a dual-image separable block-based reversible data hiding algorithm in encrypted domain for embedding secret message in base5 numeral framework is proposed here. The proposed scheme has not been suffering from underflow and overflow problem so that empowering it to embed and recover information precisely from low-intensity pixels too. This property makes our proposed methodology truly reasonable for its utilization on medical images. To prove the effectiveness of our proposed approach, experiments have been performed on different test images. The average PSNR value is 54.10 dB for an embedding capacity of 327,680 bits for all test images which demonstrates that the method is capable of giving good quality stego images even at high payload also. The experimental study revealed that for all types of test images, the proposed methodology altogether beaten all the compared methodologies in its ability to embed secret message and precisely recover it by maintaining the visual quality of stego images too.

Quadratic surface support vector machine with L1 norm regularization

&lt;p style='text-indent:20px;'&gt;We propose &lt;inline-formula&gt;&lt;tex-math id="M1"&gt;\begin{document}$ \ell_1 $\end{document}&lt;/tex-math&gt;&lt;/inline-formula&gt; norm regularized quadratic surface support vector machine models for binary classification in supervised learning. We establish some desired theoretical properties, including the existence and uniqueness of the optimal solution, reduction to the standard SVMs over (almost) linearly separable data sets, and detection of true sparsity pattern over (almost) quadratically separable data sets if the penalty parameter on the &lt;inline-formula&gt;&lt;tex-math id="M2"&gt;\begin{document}$ \ell_1 $\end{document}&lt;/tex-math&gt;&lt;/inline-formula&gt; norm is large enough. We also demonstrate their promising practical efficiency by conducting various numerical experiments on both synthetic and publicly available benchmark data sets.&lt;/p&gt;

When will gradient methods converge to max‐margin classifier under ReLU models?

We study the implicit bias of gradient descent methods in solving a binary classification problem over a linearly separable data set. The classifier is described by a non‐linear ReLU model and the objective function adopts the exponential loss function. We first characterize the landscape of the loss function and show that there can exist spurious asymptotic local minima besides asymptotic global minima. We then show that gradient descent (GD) can converge to either a global or a local max‐margin direction or may diverge from the desired max‐margin direction in a general context. For stochastic gradient descent (SGD), we show that it converges in expectation to either the global or the local max‐margin direction if SGD converges. We further explore the implicit bias of these algorithms in learning a multineuron network under certain stationary conditions and show that the learned classifier maximizes the margins of each sample pattern partition under the ReLU activation.

Network Optimization via Smooth Exact Penalty Functions Enabled by Distributed Gradient Computation

This article proposes a distributed algorithm for a network of agents to solve an optimization problem with separable objective function and locally coupled constraints. Our strategy is based on reformulating the original constrained problem as the unconstrained optimization of a smooth (continuously differentiable) exact penalty function. Computing the gradient of this penalty function in a distributed way is challenging even under the separability assumptions on the original optimization problem. Our technical approach shows that the distributed computation problem for the gradient can be formulated as a system of linear algebraic equations defined by separable problem data. To solve it, we design an exponentially fast, input-to-state stable distributed algorithm that does not require the individual agent matrices to be invertible. We employ this strategy to compute the gradient of the penalty function at the current network state. Our distributed algorithmic solver for the original constrained optimization problem interconnects this estimation with the prescription of having the agents follow the resulting direction. Numerical simulations illustrate the convergence and robustness properties of the proposed algorithm.