Search

Because huge amount of scientific papers have been published at an accelerating rate, it is beneficial to do intelligent paper classification, especially fine-grained classification. However, existing natural language processing techniques are mostly coarse-grained. Some characteristics of fine-grained scientific paper classification needs special attention. One is that the number of data may well be quite limited. Number of papers in the lower level sub-fields inevitably becomes less. Meanwhile, emerging sub-fields with new discoveries will have few papers, nevertheless these sub-fields can be important. Furthermore, fine-grained labeling of scientific papers requires high expertise and is time consuming. Another aspect of scientific papers is that they contain multi-modal information. To address the above two issues, we propose a multi-modal hierarchical fusion network (MHFNet) for fine-grained paper classification. We treat paper abstract features, image features, and paper title features as three modalities. The MobileNetV2 model and the ALBERT model are combined in the proposed model to encode multi-modal information. Comparison results with baseline methods on both sufficiently large datasets and number-limited datasets show improvements, even more on number-limited datasets. [ABSTRACT FROM AUTHOR]

Computer vision is one of the artificial intelligence's most challenging fields, enabling computers to interpret, analyse and derive meaningful information from the visual world. There are various utilizations of computer vision algorithms, and most of them, from simpler to more complicated, have an object and shape recognition in common. Traditional pen and paper tests are designed in a pre-established format and consist of numerous basic shapes, which designate the important parts of the test itself. With that in mind, many computer vision applications regarding pen and paper tests arise as an opportunity. Massive courses and large schooling organizations mostly conduct their exams in paper format and assess them manually, which imposes a significant burden on the teaching staff. Any kind of automatization that will facilitate the grading process is highly desirable. Hence, an automated answer recognition system in assessment was developed to mitigate the problems above. The system uses images of scanned test pages obtained from the test scanning process and performs the necessary image manipulation steps to increase target recognition accuracy. Further, it manages to identify regions of interest containing multiple-choice questions and contours. Finally, the system verifies obtained results using the knowledge of the whereabouts of the test template regions of interest. [ABSTRACT FROM AUTHOR]

Although paper shredding is widely used to prevent confidential papers from being misused, still it cannot be considered a convenient process. The size of paper shreds became smaller and smaller, as new methods of shredded paper reassembly and reconstruction are evolving. This paper focuses on clustering, which is a possible phase in the assembly process. This work considers real strip-cut shreds, in addition to images shredded by a simulator in one direction to make strip-cut shreds of different sizes, from wide to narrow shreds, and images shredded in two directions, possibly reflecting cross-cut and micro-cut shreds. K-means is used to cluster shreds, the features tested are gray-level ranges, and the well-known gray-level co-occurrence matrix, invariant moments, segmentation-based fractal texture analysis algorithm, and color moments. The number of shreds grouped in the same cluster with originally adjacent neighbors is used to indicate clustering effectiveness, in addition to the overall accuracy of strip-cut shreds clustering. When the number of clusters is 5, and the k-means experiments run 100 times for 38 images, the overall accuracy of gray-level ranges in simulated strip-cut shreds is 84.87, 89.27, and 93.5 percent in the three different sizes tested, also in cross-cut and micro-cut shreds, gray-level ranges achieve a relatively high number of shreds with 3 and 4 originally adjacent neighbors found in the same cluster. [ABSTRACT FROM AUTHOR]

This paper focuses on the adverse impact of outlier points and the ambiguity of candidate localizations in 3D object detection in terms of point cloud dataset. First, outlier points can disperse real feature extracting and mislead object detection, we propose an outlier weakening strategy. The neighborhood points of each point in the point set can be established via multi-directional search algorithm, and the correlations among points in the neighborhood are figured out via self-attention mechanism, then each point representation can be enhanced with the key information from its neighborhood, thus the negative impact of outlier points will be weakened due to obtaining real knowledge of object from neighborhood context. Second, multiple proposed boxes for object localization usually containing the same sampling points, this causes vagueness in differing them from each other and leads to incorrect object positioning. This paper proposes a voxel coding strategy with adaptive pooling, the candidate boxes are divided into voxels, and each voxel is further divided into multiple columns, then they are weighted and aggregated according to the importance of each column, thus can pop out the most confident spatial voxel encodings as reliable object localization nominees. This algorithm achieves an average accuracy of 82.98% and 93.2% on the KITTI dataset Car category and ModelNet40 dataset, respectively. [ABSTRACT FROM AUTHOR]

Authors, editors, and reviewers need to have a good perception regarding the quality of a manuscript in order to improve their skills, save effort, and prevent errors that can affect the submission procedure. In this paper, we compared the author's perception of a manuscript's quality with the manuscript's actual impact. In addition, we analyzed the uncertainty of the author's perception of the manuscript's quality. From there, we defined 'partition' as the author's ability to perceive the actual quality. We did this by launching a website for the use of the scientific community. This webpage provided a tool to help improve an investigator's skill in understanding and recognizing the quality of a manuscript so as to help researchers improve and maximize their works' potential impact. We carried out the experiment with 106 experienced users who tested our webpage. We found that the Abstract, the Title, and the Keywords were enough to perform a substantially decent evaluation of a manuscript. Most of the researchers were able to determine the quality of a paper in less than a minute from this small amount of information. [ABSTRACT FROM AUTHOR]

In today's world, health and medicine play an undeniable role in human life. Traditional and current Electronic Health Records (EHR) systems that are used to exchange information between medical stakeholders (patients, physicians, insurance companies, pharmaceuticals, medical researchers, etc.) suffer weaknesses in terms of security and privacy due to having centralized architecture. Blockchain technology ensures the privacy and security of EHR systems thanks to the use of encryption. Moreover, due to its decentralized nature, this technology prevents central failure and central attack points. In this paper, a systematic literature review (SLR) is proposed to analyze the existing Blockchain-based approaches for improving privacy and security in electronic health systems. The research methodology, paper selection process, and the search query are explained. 51 papers returned from our search criteria published between 2018 and Dec 2022 are reviewed. The main ideas, type of Blockchain, evaluation metrics, and used tools of each selected paper are discussed in detail. Finally, future research directions, open challenges, and some issues are discussed. [ABSTRACT FROM AUTHOR]

Vehicles are getting equipped in technology, communication between vehicle and the user is getting better. The vehicular network is an emerging technology to provide mobile users with the flexibility to use various services such as entertainment and navigation on wheels. The users are served with data related to their travel in terms of road map, weather updates, traffic congestions, radio services, social network applications and place of interest. The data communication takes place with the help of Smart On-board Unit (SOBU) present on the vehicle. There is a requirement of compound data computations with firm latency. Vehicle Edge computing (VEC) is the emerging technology that has serves at the edge in the neighbourhood of vehicle that enables data offloading. There will be much of energy consumption and latency resulting from offloading and computations. In this paper smart offloading scheme is proposed that will efficiently harvest the energy and reduce the energy consumption problem. As a performance statistic, the execution cost is used, which accounts for both execution delay and task failure. And also, the offloading scheme is analyzed for single user and multiuser by simulations and the results are compared graphically for battery energy level, average execution cost and channel mode select parameters respectively. The factors affecting both Single-user and multi-user computation are identified from the results. Experimental results show that offloading schemes proposed for single user and multi user work better than the other state-of-the-art algorithms for vehicular networks. This single-user analysis will be a benchmark for developing auto pilot vehicle for people with special needs and multi-user analysis will help in developing to all kinds of Electric Vehicle communication. [ABSTRACT FROM AUTHOR]

Loop closure detection is a key component of visual SLAM(Simultaneous Localization and Mapping). However, the existing loop closure detection algorithms are easily affected by the illumination change and object change of the scene. Since semantic features of images can improve the accuracy of object location recognition, a loop closure detection algorithm based on image semantic features and bag-of-words model is proposed in this paper. Because of the evenly distributed image features can better reflect the content of the image. So firstly, the ORB feature extraction algorithm is improved to make the extracted feature points more evenly distributed in the image, and then the extracted feature points are used to build the bag-of-words model. Then the L2 norm is adopted to calculate the similarity between images, and according to which the loop closure candidate images are determined quickly. In order to reduce the adverse effects of illumination changes and object changes on loop closure detection, YOLOv4 is used to extract semantic features of images in this paper, and real loop closure will be screened from the candidate images according to cosine values of included angles between similar objects in different images, so as to complete the loop closure detection. Experiments on TUM dataset and actual images show that the proposed algorithm can effectively reduce the adverse effects of illumination changes and object changes on loop closure detection, and effectively improve the accuracy and adaptability of loop closure detection. [ABSTRACT FROM AUTHOR]

Paper folding-and-punching tests are used to measure spatial abilities in humans. This paper presents a qualitative model for paper folding (QPF) and a computer game (Paper Folding Game) developed to apply and show the reasoning capabilities of the QPF model. This interactive game presents paper-folding activities intended to help users train and understand how to fold a paper to get a specific shape. Then, it presents paper-folding-and-punching tests to the players. The Paper Folding Game can automatically generate paper-folding-and-punching questions with varying degrees of difficulty depending on the number of folds and holes made, thus producing additional levels for training. The reasoning mechanisms in the QPF model are used by the Paper Folding Game to infer the right answer to each paper-folding-and-punching question. This reasoning capability allows the game to provide feedback to the players when they are wrong and also to create other plausible answers automatically, so that random question-answers are shown to the players in the master-mode. The Paper Folding Game has been implemented using Unity engine and it is available to download from GooglePlay and AppleStore for everyone to train their spatial reasoning skills. [ABSTRACT FROM AUTHOR]

The need of keeping medical information safe and secure stems from the fact that doctors rely on it to make accurate diagnoses. If this information is altered in any way, no matter how minor, there is a risk of an inaccurate diagnosis, which could result in severe medical issues and death. The transition from paper to electronic health records (EHRs) has considerably improved patient care quality and efficiency. However, for many healthcare service providers, it has extended the attack surface. Because of the value of a patient's medical information, this has posed a threat to both patients and healthcare providers. When security is not taken into account in healthcare systems, patients' privacy is jeopardized. The intended solution to this challenge is to create a modified AES algorithm to secure patient medical information. Although, the AES algorithm is secure, however, there is always a need for improvement on any cryptographic algorithms in terms of computational cost. This study implements AES and modified the last round of the AES and their performance has been measured by scrambling input datasets of various contents and volumes. The experimental results show that modified AES outperforms AES algorithms in terms of Encryption time while AES outperform modified AES in terms of decryption time. Also, the Avalanche effect results revealed that modified AES has a higher avalanche effect for small-size files while a smaller avalanche effect for larger file sizes. This signifies that modified AES security is stronger for a small size file while conventional AES has higher security for larger file sizes. The average encryption time of the AES algorithm for text files is 1513.3ms while the modified AES average encryption time gives 1293.837ms. The average decryption time for conventional AES is 1289.627ms while the average decryption time for modified AES give 1400.136ms. Modified AES uses lesser time complexity during the encryption of all categories of data files while conventional AES uses lesser time complexity during the decryption of all categories of data files. [ABSTRACT FROM AUTHOR]

This is an active community that aims at advancing knowledge in the design, development, and application of novel methods, tools, and techniques for HCI. The special issue on "New Frontiers in Multimedia-based and Multimodal HCI" considers research in the area of multimedia-based and multimodal systems for advanced human-computer interaction (HCI) for exploring and expanding on its frontiers. The authors tested the proposed framework to evaluate an Augmented Reality app for smartphones and assess if it could be useful to stimulate interest and awareness about the specific urban area where it is used. [Extracted from the article]

Textual Semantic Similarity is a crucial part of text matching tasks, and it has a very wide range of applications in natural language processing (NLP) tasks such as search engines, question-answering systems, information retrieval, natural language inference. Although there are a variety of approaches about textual semantic similarity, many do not succeed in achieving the semantic representation of a sentence or text that represents it well, and ignore that different words serve different roles in expressing the meaning of the whole sentence in different degrees. Therefore, our paper proposes a Siamese Bert network model to obtain textual semantic similarity. Firstly, we utilize the Bert network model to obtain the semantic features of each word in the sentence as input and utilize the merit of the Siamese network, reducing the training parameters, sharing the same encoder and feature weight information with each other. Then we use the attention mechanism to obtain more advanced semantic features. Furthermore, the similarity between two sentences can be derived by the methods of calculating the distance or concatenating their high-level semantic representations. In this paper, we apply the network structure to three related semantic similarity datasets, which perform better than other approaches. [ABSTRACT FROM AUTHOR]

Mahalanobis Taguchi System (MTS) is used for pattern recognition and classification, diagnosis, and prediction of a multivariate data set. Mahalanobis Distance (MD), orthogonal array (OA), and signal-to-noise ratio (SNR) are used in traditional MTS in order to identify and optimize the variables. However, the high correlation among variables shows an effect on the inverse of the correlation matrix that uses in the calculation of MD and hence affects the accuracy of the MD. Therefore, Mahalanobis-Taguchi-Gram-Schmidt (MTGS) system is proposed in order to solve the problem of multicollinearity. The value of MD can be calculated by using the Gram-Schmidt Orthogonalization Process (GSOP). Besides, the computational speed and the accuracy in optimization using OA and SNR are other issues that are concerned the authors. Hence, the combination of MTS and other methods such as Binary Particles Swarm Optimization (BPSO) and Binary Ant Colony Optimization (NBACO) is proposed to improve the computational speed and the accuracy in optimization. The purpose of this paper is to review and summarize some works that developed and used the hybrid methodology of MTS as well as its application in several fields. Moreover, a discussion about the future work that can be done related to MTS is carried out. [ABSTRACT FROM AUTHOR]

To take advantage of the speed advantage of the Camshift and try to overcome the problem of its poor robustness in target tracking, in this paper, a real time target face tracking algorithm based on saliency detection and Camshift is proposed. Considering that the target to be tracked is more significant than the background in the frame, the saliency detection algorithm MBplus is first used to remove the background around the target as much as possible, so as to reduce the interference caused by the background to the Camshift tracking results. Then the Camshift is used to search and localize the targets in the processed video frames. At the same time, to compensate for the lack of tracking ability of Camshift for some characteristic targets, the Kalman filter is used to predict the position of the target in the current frame. Finally, the Kalman-predicted target position, the target position obtained by Camshift, are compared with the target tracked in the previous frame, and the position with high similarity is considered as the target tracking result of this paper. The experimental results show that the average tracking precision of the proposed target face tracking algorithm on the Birchfield database is 94.0%, its average tracking success rate on the NRC-IIT Facial Video Database is 100%, and even for the target faces with few attributes in ytcelebrity database, its tracking precision and tracking success rate are all 100%, which are superior to some state-of-the-art tracking algorithms. [ABSTRACT FROM AUTHOR]

Aiming at the lack of coarse-grained features in butterfly image classification and recognition research, low recognition accuracy, and limited spatial invariance to input data, this paper proposes a butterfly image classification method based on U-Net and STL-MSDNet. Firstly, in order to reduce the influence of complex backgrounds on butterfly image recognition, the U-Net model is used to segment the butterfly ecological image. Then, an STL-MSDNet model is proposed to classify butterfly images. In STL-MSDNet, Spatial Transformer Network (STN) is added to reverse the spatial transformation of butterfly images to eliminate the deformation of image butterflies and make the recognition of the classification network simpler and more efficient. Then, the Laplace pyramid (LP) was introduced to replace gaussian down-sampling in MSDNet, and the butterfly images were decomposed into different spatial frequency bands to obtain butterfly feature maps of three scales, and then they were fused to improve the feature extraction capability of the network. The experimental results show that the butterfly image semantic segmentation algorithm based on U-Net has a good effect and is suitable for the field of image segmentation in complex backgrounds. Compared with MSDNet, DenseNet and traditional classification algorithms, the butterfly image classification model based on STL-MSDNet proposed in this paper has a better effect, better robustness, and a higher recognition rate. The method proposed in this paper solved the problem of low accuracy of classification of butterfly images in complex backgrounds by existing methods, and obtains a classification accuracy of 93.8%, indicating that it has good results in the fine classification of butterflies and can be applied to butterfly identification and realize the application of butterfly ecological research. [ABSTRACT FROM AUTHOR]

In today's fast-paced world, recommendation systems have become indispensable tools, aiding users in making personalized decisions amidst an overwhelming array of choices. These systems leverage user data and preferences to generate tailor-made recommendations based on individual tastes and behaviors. This research paper introduces the development and implementation of Pranik Movies, an ultimate recommendation system for personalized movie suggestions. The system incorporates collaborative and content-based filtering techniques, utilizing machine learning algorithms to analyze user behaviors, ratings, and viewing histories. A comprehensive overview of the research framework is provided, encompassing system architecture, data pre-processing, feature engineering techniques, and model selection and design. Text processing methods such as stemming, bag-of-words (BoW), and TF-IDF (Term Frequency-Inverse Document Frequency) are employed for processing and analyzing textual movie data. The accuracy of recommendations is enhanced through the assessment of film similarities, utilizing algorithms like cosine similarity and Euclidean distance. The paper concludes by outlining future directions for advanced machine learning techniques, social media integration, expanded content support, and the refinement of the evaluation framework. Pranik Movies signifies a significant advancement in recommendation systems, enabling personalized and precise movie recommendations within a vast and diverse cinematic landscape. [ABSTRACT FROM AUTHOR]

Image dehazing techniques are widely used in complex outdoor environments and are commonly categorized based on learning mechanisms. However, the imaging mechanism reveals the reasons for the degradation of hazy images, and the imaging physics process is essential for solving clean image reconstruction. Therefore, different from the previous categorization based solely on learning mechanisms, we propose a more fundamental approach that divides the techniques based on the imaging models used and analyze the advantages and disadvantages of various imaging models to find reasonable computational methods for image reconstruction. This paper focuses on analyzing the principles of different atmospheric imaging models and discusses the dehazing methods based on these models. In addition, we also discuss the development of atmospheric scattering models and the application of different atmospheric imaging models in image dehazing. Finally, this paper presents the application effects of different atmospheric scattering models on thin fog and dense fog datasets. Various issues and challenges faced by existing image dehazing techniques are described, and further research questions are proposed. [ABSTRACT FROM AUTHOR]

In this day and age, blockchain technology has become very popular. More and more transactions have been completed through the blockchain platform. The blockchain trading platform is fast, low-cost and high security. Many companies use blockchain for online transactions. However, with the increase in transaction volume and transaction scale, malicious users (nodes) appear, and malicious nodes participate in the blockchain network to carry out improper transactions, which brings huge losses to the transaction party. This paper proposes a Blockchain transaction model based on a malicious node detection network to ensure the safety of transaction users and enable the blockchain transaction to be traded in a safe environment. Aiming at the problem of malicious nodes deliberately submitting malicious information or obtaining Bitcoin through malicious behaviors on the blockchain, a malicious node detection model (MNDM) based on a hierarchical neural network is proposed. The hierarchical network model can calculate the key attributes according to the behavior of the nodes to detect abnormal nodes and kick them out of the blockchain system. The proposed model can avoid unnecessary losses caused by malicious nodes participating in data transmission and transactions and stop losses in time. The constructed model is called a hierarchical network model because it has two significant levels and realizes the reduction of parameter volume and the calculation of key information on the levels. Comparative tests are given in this paper. The validity of the model is proved by calculating the accuracy, precision, recall rate, and F1 score of the malicious node detection model. [ABSTRACT FROM AUTHOR]

To ensure healthy lives and promoting well-being for all in the society at all ages is one of the goals of United Nations. Specially, health of elderly people plays an important factor in productivity and prosperity of any country. According to reports, there will be over two billion elderly people worldwide by 2050. Most of elderly people live independently and need some system to protect them from any kind of fall. As old people are highly susceptible to fall due to weak body structure as well as some external conditions, researchers from academia and industries are developing fall detections systems (FDS) or devices to prevent them from fall. Hence, this paper majorly aims to review the papers on fall detection systems (FDS) to protect elderly people from any kind of fall. Papers selected for this study spans from 2017- 2023. FDS will be helpful to sustain the health of elderly persons. In view of strengthening research in this domain, this study gives an integrated and a critical review of work done in this area for both wearable, non-wearable systems and hybrid systems with research directions as the advent of new technologies like deep learning, computer vision, Internet of Things (IoT) and big data may improve the existing approaches/systems. [ABSTRACT FROM AUTHOR]

In the non-deep learning-based salient object detection methods known so far, the detection effect and robustness based on the background detection method are good. However, results are not desirable in small objects and complex scene images. This paper proposes a salient object detection algorithm, which employs a fusion framework to fuse background and frequency domain features to improve the accuracy of salient object detection. First, an improved background model is proposed for salient object detection to extract the background feature of the image. Simultaneously, the frequency domain features are obtained by the proposed frequency domain algorithm, which combines global information and local details by the Gaussian pyramid algorithm and different filters. Then, within our fusion framework, the fusion operations are guided by the self-attention mechanism to fuse background and multi-scale frequency domain features to obtain the self-attention maps. Finally, this paper introduces a fusion algorithm to derive the final saliency map from the self-attention maps. The results demonstrate that the proposed method consistently outperforms state-of-the-art approaches in four evaluation metrics on six challenging and complicated datasets and improves the accuracy of salient object detection in complex and small object scene images. [ABSTRACT FROM AUTHOR]

This paper investigated the Bitcoin blockchain and other crypto currencies confidential transaction techniques for cryptographic commitment in a zero-knowledge proof system. A cryptocurrency transaction on a blockchain contains information about the sender and the receiver transaction is public. Therefore, in order to handle our sensitive data like transaction amounts should be hidden. In this paper, we study a new novel quantum non-interactive zero-knowledge proof (QNIZK) for Quantum Range Proof (QRP). Our design of both components improves the efficiency and quantum circuit without requiring a trusted setup over prior work. Since quantum range proof is more efficient in performing q-bit measurements and explains the algorithm as well as quantum state measurement. Quantum Range Proof requires an inner product in Hilbert complex vector space to measure wave function |ψ├⟩┤ in two different quantum states |0├⟩ or ┤ |1├⟩┤ for the verifier. Thus, proving the commitment value is in the range [0 ≤ k-1], we also describe witness is measured in terms of different quantum states in non-polynomial space complexity is 0√N verification time, if the quantum circuit is unstructured search. A typical experiment shows the verification time in milliseconds for the generation of Proof for Quantum Circuits in computing. [ABSTRACT FROM AUTHOR]

Even though temporal information matters for the quality of video saliency detection, many problems still arise/emerge in present network frameworks, such as bad performance in time-space coherence and edge continuity. In order to solve these problems, this paper proposes a full convolutional neural network, which integrates temporal differential and pixel gradient to fine tune the edges of salient targets. Considering the features of neighboring frames are highly relevant because of their proximity in location, a co-attention mechanism is used to put pixel-wise weight on the saliency probability map after features extraction with multi-scale pooling so that attention can be paid on both the edge and central of images. And the changes of pixel gradients of original images are used to recursively improve the continuity of target edges and details of central areas. In addition, residual networks are utilized to integrate information between modules, ensuring stable connections between the backbone network and modules and propagation of pixel gradient changes. In addition, a self-adjustment strategy for loss functions is presented to solve the problem of overfitting in experiments. The method presented in the paper has been tested with three available public datasets and its effectiveness has been proved after comparing with 6 other typically stat-of-the-art methods. [ABSTRACT FROM AUTHOR]

Over the past decades, rice crops have been crucially acknowledged as one of the most powerful energy streams for the production of resources. Plant phenotyping trait estimation includes the external feature evaluation of the plants for production growth. Phenotyping using machine learning techniques outperforms the other imaging techniques for the analysis of traits including leaf, seed, branch, panicle, flower root, shoot, etc. Rice plants, categorized by multiple traits such as growth analysis and disease management, are considered a contributing factor to the agricultural, economic, and communal losses in the upcoming development of the agricultural field. The last 15 years' diagnosis of plant disease in relation to image processing techniques has remained an area of interest among researchers. Several disease detections, identification, and quantification methods have been developed and applied to a wide variety of crops. This paper reviews the related research papers from the period between 2007 and 2023, with a focus on the development of the state of the art. The related studies are compared based on image segmentation, feature extraction, feature selection, and classification. This paper also outlines the current achievements, limitations, and suggestions for future research associated with the diagnosis of rice plant growth analysis and disease identification using machine learning techniques. [ABSTRACT FROM AUTHOR]

Many methods have been proposed to detect the originality of an image. One of the most commonly used method, the Copy - move forgery detection (CMFD), is considered here. The contribution of this paper is the application of the new fuzzy distances in clustering using metaheuristics. The family of the used fuzzy distances satisfies the axioms of the fuzzy metric. CMFD method, which includes Variable Neighborhood Search (VNS) and Bee Colony Optimization (BCO) metaheuristics, has been tested and compared with similar methods. The proposed method with the proposed new metric used in this research gave better results than the existing methods. The proposed fuzzy metrics in this paper as well as the problem of p - median clustering applied to the problem and compared with existing research in this field give better results. [ABSTRACT FROM AUTHOR]

In the digital environment, a ransomware detection and protection solution is crucial. Because it makes it possible for companies to combat the rising threat of ransomware attacks, prevent financial losses, preserve crucial systems, and satisfy legal requirements. The primary issue is that current ransomware detection and mitigation methods could be more effective due to its dynamic nature and an insufficient up-to-date understanding of its variants. To create more effective defenses and narrow the cybersecurity knowledge gap, interdisciplinary research that examines ransomware's coding, behavior, and goals is required. This paper seeks to improve methods of ransomware defense by analyzing the code, behavior, and goals of numerous ransomware variants. The study proposes using semantic similarity algorithms to estimate the severity of attacks and identify connections between existing attacks to enhance detection and mitigation measures. The objectives of this paper include developing an improved, all-inclusive multiclass ransomware detection and response generation model with automatic or semi-automatic response capabilities. The method combines artificial intelligence with semantic similarity detection techniques to automatically identify and classify ransomware attacks using predefined classifiers. Single machine learning techniques were initially trained on a dataset of ransomware samples; however, the accuracy was poor, ranging from 11% with the LGBM classifier to a maximum of 51% with the decision tree classifier. A hybridization of ML algorithms increases accuracy, leading to a notable improvement in classifier accuracy, with a Hybrid 3 obtaining 91% accuracy. Semantic algorithms periodically retrain the model to keep it current with new ransomware variations. These algorithms also provide the severity of the attack. If the criticality is low, the model notifies the administrator for additional analysis; if it is medium, sensitive data-related operations are stopped with automatic backups; and if it is high, all processes are controlled to prevent further harm. The key advantages of this model are its ability to identify 21 different ransomware classes, enhance training with new variants, and provide criticality recommendations for optimal decision-making during a ransomware assault. [ABSTRACT FROM AUTHOR]

In order to aim for diagnosis of disease and decision making, the medical imaging plays an important role in health science. Clinical pictures are portrayals of profoundly vulnerable computerized images, those can be tampered without leaving any visual clues. Hence it is challenging to keep up its credibility. However, as there are numerous ways to manipulate an image, correspondingly various strategies have also been proposed to safeguard the genuineness of medical images. This survey paper presents different techniques used for medical image authentication viz. watermarking, signature and hybrid techniques. The state-of-the-art techniques have attained promising results for authentication and tampering detection, but an efficient tampering localization and recovery have remained still as a challenge. This review article can be considered as a benchmark survey paper as it gives a complete comprehensive overview commencing from the evolution of medical image formats, types of medical imaging modalities and also provides an elaborative comparison over 40 research works in terms of prominent factors like type of medical image used, embedded region, embedded data, about the coverage of tampering localization and recovery along with the discussion on limitations and future works of each one. [ABSTRACT FROM AUTHOR]

In this rapidly advancing era, technology has been progressing extensively and swiftly. As a result, the emergence of numerous Cyber-Physical Systems (CPS) has become imperative to meet the technological demands of modern life. However, these systems generate a substantial amount of data, which poses challenges in terms of management, storage, and susceptibility to external attacks. This paper primarily focuses on the performance and security aspects of CPS, particularly in countering external threats, through the integration of blockchain technology and machine learning. It provides a comprehensive review of recent research findings that demonstrate the use of blockchain to enhance CPS performance while ensuring robust security. Furthermore, the paper explores the synergistic application of blockchain and machine learning techniques to reinforce CPS security. Moreover, it investigates how the combination of blockchain with physically unclonable functions (PUF) can significantly enhance the efficacy of physical device authentication. [ABSTRACT FROM AUTHOR]

Test anxiety is an emotional state characterized by subjective feelings of discomfort, fear, and worry that can considerably affect students' academic performance. Virtual Reality exposure (VRE) is a promising approach to address test anxiety, but the few VRE systems for test anxiety in the literature concern only written exams. Since oral exams elicit more anxiety than written exams, the availability of VRE systems for oral exams would be precious to a large population of students worldwide. Another limitation of existing VRE systems for test anxiety is that they require the availability of a head-mounted display, posing a barrier to widespread use. This paper aims to address both issues, proposing a VRE system that deals with oral exams and can be used with common PC displays. The design of the proposed system is organized in three oral test scenarios in which a virtual agent acts as the student's examiner. The virtual examiner behaves friendly in the first scenario and increasingly reduces its friendliness in the two subsequent scenarios. The paper assesses the feasibility for VRE of the proposed system with two complementary methods. First, we describe a quantitative user study of the three system scenarios, showing that they induce increasing levels of anxiety. Second, we present a qualitative thematic analysis of participants' post-exposure interviews that sheds further light on the aspects of the virtual experience that contributed to eliciting negative or positive affect in participants, and provides insights for improving VRE systems for test anxiety. [ABSTRACT FROM AUTHOR]

Developing offline reinforcement learning evaluation applications faces challenges such as heterogeneous data and algorithm integration, user-friendly interface, and flexible resource management. This paper designs and implements ORLEP, an efficient platform to provide high-level services for offline reinforcement learning evaluation. Besides integrating underlying infrastructure with highly concurrency and reliability, core components with distributed deployment and 3rd party libs and benchmarks incorporation, ORLEP supplies high-level abstractions for (1) data management, (2) model training and evaluation, (3) result visualization, and (4) resource configuration and supervision. Moreover, this paper verifies specific cases and the results demonstrate the performance and scalability of the proposed ORLEP. [ABSTRACT FROM AUTHOR]

For the problem of blurred images in the process of UAV video inspection of oil and gas pipelines, a PSO-2D-VMD-BD denoising algorithm is proposed in this paper. The procedure of the developed algorithm is that: (1) the particle swarm optimization (PSO) is used to optimize K and α of the two-dimensional variational modal decomposition (2D-VMD); (2) use the 2D-VMD to decompose the oil and gas pipelines image into K intrinsic mode functions (IMFs); (3) use the Bhattacharyy Distance (BD) and the probability density function (PDF) to distinguish the signal-dominant IMFs (S-D IMFs) and the noise-dominant IMFs (N-D IMFs), and then use the wavelet threshold denoising method to eliminate the noise in the N-D IMFs; and (4) finally, the oil and gas pipelines image is reconstructed using the denoised N-D IMFs and S-D IMFs to obtain the denoised image. Compared to median filtering, wavelet threshold filtering, the Rudin Osher Fatemi (ROF), the 2D-VMD-BD, and the PSO-2D-VMD-KLD, theoretical analysis and simulation results indicate that the solution presented in this paper achieved a decent denoising effect, which can provide a solid foundation for the subsequent recognition of oil and gas pipelines by UAV inspection. [ABSTRACT FROM AUTHOR]

This paper examines and assesses state-of-the-art proposed machine and deep learning techniques for breast cancer identification and classification based on breast screening image modalities. Ten research questions related to the medical image modalities, image dataset, image pre-processing, segmentation and classification techniques are framed to identify the scope of the review. From the perspective of research questions, an extensive review is carried out with various research papers, book chapters published in SCI/Scopus-indexed journals and international conferences from 2010 to 2021. Many issues such as image modalities, segmentation techniques, features, and evaluation metrics are identified with the machine and deep learning methodologies. This review shows that about 57% of the selected studies have used digital mammograms for breast cancer identification and classification. Most of the selected studies have used public datasets and employed noise removal, data augmentation, scaling, and image normalization techniques to alleviate the inconsistencies in breast cancer images. It is observed that mainly thresholding-based, region-based, edge-based, clustering-based, and deep learning (DL) based segmentation techniques are used in many studies. It has also been observed that the support vector machine (SVM) and variants of convolutional neural network (CNN) are the most used classifiers for breast cancer identification and classification. It is found that CNN-based classification models have achieved an accuracy of 100% in the classification of breast cancer for 250 ultrasound images. This review may help researchers to figure out whether a machine or deep learning technique works better on a particular dataset and which features are significant for breast cancer detection. Traditional machine-learning approaches are mostly used in classification, whereas deep-learning techniques have conquered the field of image analysis. This review presents the strengths and weaknesses of the existing machine and deep learning-based models. This review is summarized by providing appropriate answers to the formed research questions with future recommendations in the identification and classification of breast cancer. Highlights: Illustrate the applications of the machine and deep learning techniques in breast cancer detection. Illustrate the medical image modalities used to identify and classify breast cancer. Present the breast image datasets used in the classification models for medical images. Illustrate the image pre-processing, image segmentation, feature extraction techniques and classification algorithms. Provide machine learning techniques for breast cancer identification and classification using various medical image modalities. Provide deep learning techniques for breast cancer identification and classification using various medical image modalities. Identify research gaps and recommendations for the future are proposed. [ABSTRACT FROM AUTHOR]

YouTube video recommendation algorithm plays an important role in enhancing user engagement and profitability (or monetization), yet it struggles to deal with disturbing visual content. Real-time analysis using deep learning techniques can play a vital role to identify and filter the disturbing content in online videos. In this paper, we propose an end-to-end trainable two-stream deep learning framework that analyzes and classifies the different categories of disturbing content embedded in child-friendly cartoon videos on YouTube and YouTube Kids platforms. At first, the model extracts the static and motion features from videos through an individual pretrained convolutional neural network (CNN) i.e., EfficientNet-B7. In the next phase, the extracted features are processed by spatio-temporal bidirectional long short-term memory (BiLSTM) network to capture the long-term global temporal dependencies of videos. The learned video representations are forwarded to the classifier for multiclass classification into six categories. Three different types of fusion strategies are investigated to combine the spatial and temporal streams. The evaluation of these methods is performed through extensive experiments on a customized large-scale dataset–YouTube cartoon content filtering (YT-C2F) dataset. The spatio-temporal EfficientNet-BiLSTM network with feature-level fusion displays the best results (f1-score = 0.9316) and shows the efficiency of using a two-stream network compared to single-stream baseline methods. This paper makes two-fold contributions. First, using only static information from videos in a deep learning framework is less effective than using both static and motion information. Second, the feature-level fusion of two streams of networks generates better classification results than early and late fusion techniques. The performance comparison of the proposed model with existing state-of-the-art techniques confirmed the competitive or even better results of our framework. [ABSTRACT FROM AUTHOR]

The exponential growth and wide-area applications of the Internet of Things have garnered a lot of interest from academics and industries, thus becoming one of the most actively studied research paradigms in recent years. Internet of Things is a concept that builds an inter-connected environment where physical devices collect data from the surroundings and customize it to our requirements. It is known for its fast processing and quick response. Edge computing is a platform that performs the computations on the data supplied by the Internet of Thing devices at the network edge that leads to real-time processing. Despite its advantages, privacy protection and security challenges remain a critical concern that must be addressed. This paper aims to give a comprehensive review on cryptographic schemes used for securing edge computing. In particular, we first present a concept of edge computing in the context of IoT including architecture and advantages over cloud computing. We have explored various encryption techniques like identity-based encryption, attribute-based encryption, searchable encryption, and homomorphic encryption that secure the sensitive data before processing it at the network edge. Various parameters of each encryption technique, such as the length of the public key, private key, and ciphetext, are compared. Computational complexity of encryption and decryption is also included in our comparative study. Although many review papers focused on authentication and authorization challenges, this review on encryption techniques in context of edge security was highly needed. We have also looked at some of the other review papers and compared them to our survey to show how important it is. [ABSTRACT FROM AUTHOR]

Fully capturing valid complementary information in multi-view data enhances the connection between similar data points and weakens the correlation between different data point categories. In this paper, we propose a new multi-view clustering via dual-norm and Hilbert-Schmidt independence criterion (HSIC) induction (MCDHSIC) approach, which can enhance the complementarity, reduce the redundancy between multi-view representations, and improve the accuracy of the clustering results. This model uses the HSIC as the diversity regularization term to capture the nonlinear relationship between different views. In addition, l1-norm and Frobenius norm constraints are imposed to obtain a subspace representation matrix with inter-class sparsity and intra-class consistency. Moreover, we also designed a valuable approach to optimizing the proposed model and theoretically analyzing the convergence of the MCDHSIC method. The results of extensive experiments conducted on five challenging data sets show that the proposed method objectively achieves a highly competent performance compared with several other state-of-the-art multi-view clustering methods. [ABSTRACT FROM AUTHOR]

In the digital era, data transmission plays a crucial role, encompassing various forms such as text, images, and videos. Industries such as banking, medical, marketing, social media, and business applications heavily rely on images for communication with customers and others. However, this increased reliance on digital communication has attracted Cyber Attackers (CAs), individuals or entities that exploit network weaknesses and disturb data integrity, leading to vulnerabilities in data transmission. The healthcare industry, in particular, has faced significant costs due to data breaches, as indicated by the Cost of a Data Breach Report 2021. Securing such data during communication is a key challenge. Visual Cryptography (VC) emerges as a powerful solution for secure image transfer. VC's share generation module encrypts the image and divides it into multiple shares, which are then communicated to recipients. Individual shares do not disclose classified information. At the destination, the shares are digitally or physically stacked to reconstruct the original image without the need for complex mathematical operations or additional hardware. Key constraints in VC include the security of share images, the quality of the reconstructed image, pixel expansion issues, and computational complexity. Despite the significance of VC, existing research papers often focus on specific types of schemes, leaving gaps in comprehensive comparisons and coverage of the state-of-the-art. To address this, we propose a new Progressive Meaningful Visual Cryptography (PMVC) method in this paper for transferring secret images by triggering an error instance. The proposed method generates meaningful shares and ensures the quality of the reconstructed image, achieving a peak signal-to-noise ratio (PSNR) of up to 37 dB without pixel expansion issues. Additionally, the proposed method works on (n, n) VC schemes and supports grayscale secret images. [ABSTRACT FROM AUTHOR]

Citation count prediction is a subfield of bibliometrics, which involves using mathematical and statistical models to predict the quantity of citations that a scholarly research paper will accept in the future. This study suggests a novel approach for forecasting the number of citations academic papers will receive in future. The proposed model includes four major phases: "(a) pre-processing, (b) feature extraction, (c) feature selection, (d) Three-Layer-Deep-Learning-based-citation-count-prediction". In this paper, the Raw data collected is pre-processed via stop word removal, stemming, lemmatization, tokenization. As a resultant, the tokens are acquired. The features are extracted from the tokens. The features like bag-of-words, Part-of-Speech (POS) Tagger, Pointwise Mutual Information (PMI), Class distribution factor-based Term Frequency-Inverse Document Frequency (CDF-TFIDF), Doc2vec (document to vector) and Weighted Latent Semantic Analysis (WLSA). Subsequently, from the extracted features, the optimal features are selected using the new hybrid optimization model. The proposed hybrid optimization model- Polar Red Fox Optimization (PRFO), is the agglomeration of concepts of the standard "Polar Bear Optimization (PBO)" and "Red the fox optimization (RFO)," respectively. The citation counts are predicted using the new three-layer-deep-learning paradigm that includes the "Long Short-Term Memory (LSTM), Recurrent Neural Net Language Model (RNNL), and Convolutional Neural Network (CNN)," respectively. The input to LSTM and RNNL is the selected optimal features. The outcome from LSTM and RNNL is fed as input to the CNN. The outcome is acquired from CNN. Finally, the proposed prototypical is validated over current models, to manifest its efficiency in accurate citation count prediction. [ABSTRACT FROM AUTHOR]

With the maturation of computer-aided technology, many famous acoustic simulation software have appeared around the world. Researchers can use the software to evaluate the sound quality of the performance spaces. However, most popular software require high levels of professionalism for operators, and 3D model visualization is not intuitive enough. Aiming at these problems, this paper designs and implements an acoustic simulation system for performance space based on Unity 3D and C++. According to the system's functional design, it can be divided into two modules: the model configuration module and the acoustical parameter prediction module. We develop various modeling tools in the model configuration module and use modified image source method to simulate room impulse response in the acoustical parameter prediction module. Then, acoustical parameters such as reverberation time, clarity, definition, and speech transmission index are computed. We take Hengdian theater as an example of experimental test, and compare the results with EASE simulation results and measurements. The system test shows that the acoustic simulation system has better visualization and is easier to operate. And in this case, the absolute errors of the simulation values of each acoustical parameter are almost closer to the measurements than those of EASE, which proves that the simulation values of the system in this paper are more accurate to a certain extent. [ABSTRACT FROM AUTHOR]

China has entered a stage of high-quality development, and people have higher demand and expectations for the existing medical field. MNER(Medical Named Entity Recognition) is the task of identifying the correct entity boundary and classifying medical entities from a piece of medical text information. The effect of MNER would directly affect the performance of downstream relationship extraction and intelligent question answering, which has important research significance and value. In this paper, aiming at the named entity recognition of discontinuous medical entities in Chinese medical text information, this paper expands the research based on BiLSTM-CRF, introduces the IDCNN layer after the input layer of the model to capture the local context information in the medical text, and then uses the output of IDCNN as the input of BiLSTM-CRF for subsequent training, to construct an IDCNN-BiLSTM-CRF network model for discontinuous medical text. Based on BERT, the weight is assigned to the 12-layer transformer in BERT, and the results are weighted and averaged, and a WfBERT-Att-D-BiLSTM-CRF model based on the weight output of different layers of BERT is proposed. During the experiment, the hidden layer, the number of iterations, and the batch data size are continuously experimented with, and finally, the optimal parameter settings are obtained. In this paper, repeated experiments are carried out on different datasets, the final actual recognition effect is also tested, and the correctness of the proposed model is verified from different perspectives. [ABSTRACT FROM AUTHOR]

Accurate segmentation of rectal tumor lesion regions can provide an essential basis for clinical treatment and prognosis monitoring of tumors. However, there are many problems in the rectal tumor segmentation task at present: the lack of high-quality datasets; the mainstream segmentation network cannot complete the high-precision segmentation task of rectal tumors. In this paper, we investigate image enhancement and segmentation algorithms for convolutional neural networks and construct a rectal tumor MRI dataset by improving the CycleGAN network and loss function to achieve domain migration and reconstruction of rectal tumor CT and MRI images, given the small amount of rectal tumor image data and the existence of different modalities and regimes in CT and MRI. For the rectal tumor segmentation problem, a novel segmentation network DCMSG-UNet was designed based on the U-Net network. this network uses dilated convolution and multi-headed self-attention mechanisms to improve the base feature extraction module of the segmentation network, adds a decoder path, and uses the GAM hybrid attention mechanism to amplify the dimensional interaction features of the additional decoder path. Comparison experiments with six network models, DeepLabV3+, U-Net, UNet++, UNet3+, TransUNet, and Swin-Unet, show that the segmented region obtained from the DCMSG-UNet model proposed in this paper is closer to the real tumor region, with a DICE metric of 0.8416 and a Hausdorff distance of 11.3229, which can effectively segment the tumor. The experimental results show that our proposed method performs significantly better than the above methods, with a DICE metrics improvement of about 6%. To visualize the segmentation results, this paper designed a rectal tumor MRI image segmentation system based on PyQt5 to realize human-computer interaction and assist doctors in clinical diagnosis. [ABSTRACT FROM AUTHOR]

Multireceiver synthetic aperture sonar (SAS) can produce high-resolution images by coherently superposing successive echoed signal. However, multireceiver SAS imaging is a difficult issue since the system can be divided into many bistatic SASs. Traditional imaging algorithms cannot be directly used to process the multireceiver SAS echoed signal, as they are just applied to monostatic sonar system. In order to solve this issue, this paper proposes an imaging algorithm based on Loffeld's bistatic formula (LBF) for multireceiver SAS system. With the LBF method, the point target reference spectrum (PTRS) of multireceiver SAS system can be split into quasi-monostatic term and the bistatic deformation term. The quasi-monostatic term is similar to the spectrum of monostatic SAS while the bistatic deformation term is caused by the displacement of transmitter and receiver. In this paper, the data of multireceiver SAS is firstly coerced into monostatic one by compensating the bistatic deformation term. After this step, the chirp scaling (CS) algorithm based on monostatic SAS can be directly applied. At last, simulations are exploited to demonstrate the presented method. Compared to traditional multireceiver SAS CS algorithms, the presented method significantly improves the imaging performance. Both ghosts target and higher sidelobes that traditional multireceiver SAS CS algorithms suffer from can be successfully suppressed. Furthermore, the presented method can obtain high-resolution results which are nearly identical to the benchmark of BP algorithm. [ABSTRACT FROM AUTHOR]

Cognitive radio is providing the solution to the challenges of spectrum scarcity. The central idea behind this technology is to use spectrum without interfering with the rights of primary users, allowing multiple users to use the spectrum at the same time. Thus, the goals are to avoid high-cost spectrum resetting and improving overall spectrum utilization. This technology brings the TCP/IP protocol stack's Physical, Data Link, and Network Layers up to date with reference to Cognitive radio technology. Furthermore, due to the dynamic nature of cognitive radio Networks, these can be easily physically or mentally hurt, influenced, or attacked: to a slew of new security flaws. We concentrated on various attacks that target the Physical, Medium Access Control, and Network layers in this paper. A comparison of current defenses against outside and inside attacks is also included, as is a discussion of the current detection mechanisms and their countermeasures. Moreover, the systematic review is conducted to look into various attack vectors or malicious activities that may degrade the performance of cognitive radio networks. In addition to this, various paper discusses various recent tools and techniques that can be used to diagnose potential threats on wireless ecosystem. This survey also highlights the fundamental security challenges for cognitive radio networks. [ABSTRACT FROM AUTHOR]

In recent years, the Internet has had a main and important contribution to human life and the amount of data on the World Wide Web such as books, movies, videos and, etc. increase rapidly. Recommender systems allow users to quickly access items that are closer to their interests. One of the most popular and easiest models of recommender systems is the Collaborative Filtering (CF) model, which uses the items ratings given by users. The important challenge of CF is robust against the attacks which manipulated by fake users to reduce the efficiency of the system. Therefore, the impact of attacks on the item recommendations will increase and fake items will be easily recommended to users. The purpose of this paper is to design a robust CF recommender system, T&TRS, Time and Trust Recommender System, against user attacks. Our proposed system improves the performance of users clustering for detecting the fake users based on a novel community detection algorithm that is introduced in this paper. Our proposed system calculates the reliably value for all items ratings and tags them as suspicious or correct. T&TRS considered the rating time and implicit and explicit trust among users for constructing the weighted user-user network and detects communities as the nearest neighbors of the users to predict unknown items ratings. After detecting the suspect users and items using a novel community detection method, our proposed system removes them from rating matrix and predict the rating of unobserved items and generate the Top@k items according to the user interests. We inject the random and average attacks into the Epinion data set and evaluate our proposed systems based on Precision, Recall, F1, MAE, RMSE, and RC measures before and after attacks. The experimental results indicated that the precision of items recommendations increase after attack detection and show the effectiveness of T&TRS in comparison to the two base K-means methods such as KMCF-U, KMCF-I and graph-based methods such as TRACCF, and TOTAR. [ABSTRACT FROM AUTHOR]

Calendar systems adopted across the world are either solar, lunar or lunisolar, based on the movements of the sun, the moon, or both. The Gregorian (solar) calendars are considered as time references for modern computations. However, Vedic calendars, being lunisolar, can be more effective for the analysis of activities that depend on both celestial bodies. In this paper, we present VedicDateTime, an open-source framework that implements the Vedic calendar and provides conversions for Gregorian dates. Along with package details, we also provide two case studies that make use of the proposed package for time-series analysis. The objective of this paper is to motivate researchers to explore the potential of the Vedic calendar from the perspective of time series analysis. [ABSTRACT FROM AUTHOR]

Food plays an important role in our lives that goes beyond mere sustenance. Food affects behavior, mood, and social life. It has recently become an important focus of multimedia and social media applications. The rapid increase of available image data and the fast evolution of artificial intelligence, paired with a raised awareness of people's nutritional habits, have recently led to an emerging field attracting significant attention, called food computing, aimed at performing automatic food analysis. Food computing benefits from technologies based on modern machine learning techniques, including deep learning, deep convolutional neural networks, and transfer learning. These technologies are broadly used to address emerging problems and challenges in food-related topics, such as food recognition, classification, detection, estimation of calories and food quality, dietary assessment, food recommendation, etc. However, the specific characteristics of food image data, like visual heterogeneity, make the food classification task particularly challenging. To give an overview of the state of the art in the field, we surveyed the most recent machine learning and deep learning technologies used for food classification with a particular focus on data aspects. We collected and reviewed more than 100 papers related to the usage of machine learning and deep learning for food computing tasks. We analyze their performance on publicly available state-of-art food data sets and their potential for usage in multimedia food-related applications for various needs (communication, leisure, tourism, blogging, reverse engineering, etc.). In this paper, we perform an extensive review and categorization of available data sets: to this end, we developed and released an open web resource in which the most recent existing food data sets are collected and mapped to the corresponding geographical regions. Although artificial intelligence methods can be considered mature enough to be used in basic food classification tasks, our analysis of the state-of-the-art reveals that challenges related to the application of this technology need to be addressed. These challenges include, among others: poor representation of regional gastronomy, incorporation of adaptive learning schemes, and reverse engineering for automatic food creation and replication. [ABSTRACT FROM AUTHOR]

Security of the information stored in android-based smartphones is a hot research topic. Image steganography can be employed to increase the security of the confidential data. Images are available in various image formats. Performance of image steganography can get affected by the choice of the image format chosen owing to their different characteristics. However, the analysis of image formats has been neglected in the existing studies. Therefore, this research paper attempts to analyze the behavior and impact of various image formats (natively supported by android smartphones) on image steganography which led to better security. The objective of this research paper is threefold. Firstly, it compares and contrasts various image steganography algorithms and identifies the most appropriate for this research study. Secondly, it discusses the characteristics of image formats supported by Android smartphones and narrows down the characteristics which can affect the efficiency of image steganography. Thirdly, the research paper applies the identified algorithm on grayscale and colored images and evaluates using two different datasets. It then uses MSE and PSNR metrics to compare original images and stego images. [ABSTRACT FROM AUTHOR]

Big data analytics is an emerging topic in academic and industrial engineering fields, where the large-scale data issue is the most attractive challenge. It is crucial to design an effective large-scale data processing model to handle big data. In this paper, we aim to improve the accuracy of the classification task and reduce the execution time for large-scale data within a small cluster. In order to overcome these challenges, this paper presents a novel ensemble-based paradigm that consists of the procedure of splitting large-scale data files and developing ensemble models. Two different splitting methods are first developed to partition large-scale data into small data blocks without overlapping. Then we propose two ensemble-based methods with high accuracy and less execution time: bagging-based and boosting-based methods. Finally, the proposed paradigm can be implemented by four predictive models, which are combinations of two splitting methods and two ensemble-based methods. A series of persuasive experiments was conducted to evaluate the effectiveness of the proposed paradigm with four different combinations. Overall, the proposed paradigm with boosting-based is the best in terms of the accuracy metric compared with existing methods. In addition, boosting-based methods achieve 91.6% accuracy compared with 52% accuracy of base line model for a big data file with 10 million samples. However, the paradigm with bagging-based takes the least execution time to yield results. This paper also reveals the effectiveness of the computing Spark cluster for large-scale data and points out the weakness of RDD (Resilient Distributed dataset). [ABSTRACT FROM AUTHOR]

Image Compressive Sensing (CS) provides a scheme of low-complex image coding, but coping with the recovery quality has been a challenge. Even the excessive investment of computations into recovery cannot prevent the quality degradation due to the lack of appropriate allocation for sampling resources. In light of this, this paper fuses a context-based allocation into image CS in order to improve the recovery quality with fewer computations. Independent of original pixels, the context features of blocks are extracted from random CS samples. According to the block-based distribution on context features, more CS samples are allocated to non-sparse regions and fewer to sparse regions. The proposed context-based allocation enables a linear recovery model to accurately recover images. The contributions of this paper include: (1) an adaptive allocation involving the context features extracted from CS samples, (2) a padding Differential Pulse Code Modulation (DPCM) to quantize the adaptive CS samples, and (3) a regrouping module to improve the quality of linear recovery. Experimental results show the proposed image CS system objectively and subjectively improves the recovery quality of an image while guaranteeing a low computational complexity, e.g., it achieves average 30.85 dB PSNR value on the five 512 × 512 test images, and costs about 10 seconds on a computer with 3.30 GHz CPU and 8 GB RAM. Besides, the proposed system presents a competitive performance to the recent deep-learned image CS systems. [ABSTRACT FROM AUTHOR]

Plant disease control is an indispensable research topic in the field of agriculture. Different apple leaf diseases may have similar manifestations, and it is time-consuming and laborious to rely on manual means. In this paper, we propose an apple leaf disease classification algorithm for a small number of samples, which is based on the C-Grabcut image segmentation algorithm proposed in this paper and the improved EfficientNetB4 transfer learning algorithm. Firstly, data augmentation is used to expand the samples, which effectively solves the problems of insufficient samples and unbalanced sample categories. Then the leaves are extracted from the images using the C-Grabcut algorithm to reduce the interference brought by the background. Finally, the improved Vgg16, ResNet50, EfficientNetB0, EfficientNetB4 and EfficientNetB7 transfer learning algorithms are used to classify leaves into four categories: rust, scab, multiply and healthy. The experimental results show that the improved EfficientB4 algorithm works best with an average accuracy of 98% and the Kappa value of 0.98. In addition, the C-Grabcut algorithm reduces the training time from 153 to 73 s during an epoch, allowing the proposed algorithms to be deployed on devices with lower computing power and memory. [ABSTRACT FROM AUTHOR]

In agricultural pest management, the traditional insect population tracking in the case of several insect types is based on outsourced sticky paper traps that are checked periodically by a human operator. However, with the aid of the Internet of Things technology and machine learning, this type of manual monitoring can be automated. Even though great progress has been made in the field of insect pest detector models, the lack of sufficient amount of remote sensed trap images prevents their practical application. Beyond the lack of sufficient data, another issue is the large discrepancy between manually taken and remote sensed trap images (different illumination, quality, background, etc.). In order to improve those problems, this paper proposes three previously unused data augmentation approaches (gamma correction, bilateral filtering, and bit-plate slicing) which artificially enrich the training data and through this increase the generalization capability of deep object detectors on remote sensed trap images. Even with the application of the widely used geometric and texture-based augmentation techniques, the proposed methods can further increase the efficiency of object detector models. To demonstrate their efficiency, we used the Faster Region-based Convolutional Neural Network (R-CNN) and the You Look Only Once version 5 (YOLOv5) object detectors which have been trained on a small set of high-resolution, manually taken trap images while the test set consists of remote sensed images. The experimental results showed that the mean average precision (mAP) of the reference models significantly improved while in some cases their counting error was reduced to a third. [ABSTRACT FROM AUTHOR]

Social networking is becoming increasingly prevalent in today's globe. Young folks, senior citizens, and the general public use social media. However, over usage of social network communication harms human existence. In recent years, many mental diseases, such as reliance on cybernetic interactions, information overload, and network restriction, have been noticed in social networks. Currently, the indications of those mental diseases can be seen, posing several dangers. Human emotions, such as despair, are interior feelings that reveal an individual's actual behavior. Monitoring and detecting these thoughts from individual's choices in virtual social interactions could be very beneficial in realizing their behavior. This research develops a system that uses social media posts to detect individuals' depression. Data mining techniques have been employed to analyze social media data to identify individuals at risk of depression. It provides an overview of the major contributions made in the field of depression detection based on social networking sites using several machine learning (ML), deep learning, and data mining. We discuss the use of sentiment analysis (SA), natural language processing (NLP), ML, user profiling, and early warning systems. The paper highlights the potential benefits of using social media data for depression detection, including the ability to provide early intervention to individuals at risk of depression. However, there are also ethical considerations to be addressed, such as ensuring data privacy and avoiding stigmatization. The bag of words (BoW) approach is effective in capturing the sentiment and emotional tone of social media posts, which is crucial for detecting depression. It employs Naïve Bayes (NB) classifier which is is known for its accuracy in classifying text data for detecting the depression level. The implemented approach is compared with other SA methods concerning performance and results. In many evaluation parameters, the proposed method outperforms other SA systems. The approach provides runtime provision for adding new keywords and sentiments which adapts the approach to work fine for newest trends, wordings, and formats of posts. It suggests actions to be undertaken in case of detection of depressed users to control them drowning in further stress. Overall, this paper aims to provide a comprehensive understanding of the current state of research in depression detection based on social networking sites using data mining, and the potential implications for mental health care. [ABSTRACT FROM AUTHOR]