The application of artificial intelligence to domain-specific textual analysis has opened new possibilities in natural language processing, particularly in the context of fatwas-authoritative Islamic legal opinions. This study presents a comparative evaluation of three language embedding models-Sentence Transformers, AraBERTv2, and MARBERT-within a Retrieval-Augmented Generation (RAG) framework powered by the Gemini model. A curated corpus of 285 texts, comprising fatwas, Qur’anic exegesis, and related jurisprudential writings, serves as the basis for evaluating semantic comprehension, retrieval accuracy, and generative performance. The findings demonstrate the superior performance of Arabic-specific models, with MARBERT and AraBERTv2 notably outperforming the multilingual Sentence Transformers in capturing the nuanced language and legal reasoning typical of fatwas. These results highlight the significance of culturally and linguistically specialized models in processing religious legal discourse.

Software defect prediction is a critical task in software engineering, enabling organizations to proactively identify and address potential issues in software systems, thereby improving quality and reducing costs. In this study, we evaluated and compared various machine learning models, including logistic regression (LR), random forest (RF), support vector machines (SVMs), convolutional neural networks (CNNs), and eXtreme Gradient Boosting (XGBoost), for software defect prediction using a combination of diverse datasets. The models were trained and tested on preprocessed and feature‐selected data, followed by optimization through hyperparameter tuning. Performance evaluation metrics were employed to analyze the results comprehensively, including classification reports, confusion matrices, receiver operating characteristic–area under the curve (ROC‐AUC) curves, precision–recall curves, and cumulative gain charts. The results revealed that XGBoost consistently outperformed other models, achieving the highest accuracy, precision, recall, and AUC scores across all metrics. This indicates its robustness and suitability for predicting software defects in real‐world applications.

Deep learning has transformed image classification tasks across many domains, including medical diagnostics. Medicine wrappers, boxes, and strips often contain valuable but complex information that can be difficult to read and comprehend manually. This complexity drives users to seek additional knowledge online. However, traditional search engines often present a large volume of results, requiring users to manually filter through multiple links to find relevant information, which can be time-consuming. To address this issue, we propose a lightweight pipeline that leverages MobileNetV2 for image classification and Optical Character Recognition (OCR) for extracting text content from medicine packaging. The extracted text is processed using the RAKE algorithm to identify significant keywords, which are then matched with relevant URLs through a Google Search API. To ensure relevance, retrieved links are ranked using ROUGE scores. Performance metrics demonstrate the model's efficiency, with ROUGE-1 achieving 90% Recall, 95% F1-Score, and 90% Accuracy, and ROUGE-L achieving 83% Recall, 91% F1-Score, and 83% Accuracy. The pipeline was trained and validated on a curated dataset of 3,000 real-world medicine packaging images, publicly available on GitHub. These results highlight the novelty and practicality of our solution for automating medical information retrieval from packaging, using an interpretable and scalable deep learning-driven approach.

Recent advancements in renewable energy technologies, along with the energy exchange capabilities of Electric Vehicles (EVs), present new opportunities for enhancing renewable energy management and its integration into traditional power systems. Despite these advancements, challenges such as the pricing of charging and discharging clean energy, and the distribution of available energy supplies persist in the realm of energy trading. Our research addresses these issues by leveraging Vehicle-to-Everything (V2X) technologies, which enable EVs to distribute energy to a wide range of consumers. We introduce a dual-level optimization approach that synchronizes financial incentives with the variable electricity prices at EV charging in the parking. This approach is supported by a state-of-the-art reinforcement learning model that integrates primal-dual optimization with upper-confidence bound techniques. Our model is specifically designed to optimize both power management and the effective use of incentives. The overarching goal of this strategy is to augment the adaptability of Vehicle-to-Grid (V2G) and Vehicle-to-Vehicle systems to encourage user participation in energy exchange processes, thereby promoting a more efficient and integrated renewable energy ecosystem.

Abstract The research addresses energy consumption, latency, and network reliability challenges in wireless sensor network communication, especially in military security applications. A multi-agent context-aware model employing the belief-desire-intention (BDI) reasoning mechanism is proposed. This model utilizes a semantic knowledge-based intelligent reasoning network to monitor suspicious activities within a prohibited zone, generating alerts. Additionally, a BDI intelligent multi-level data transmission routing algorithm is proposed to optimize energy consumption constraints and enhance energy-awareness among nodes. The energy optimization analysis involves the Energy Percent Dataset, showcasing the efficiency of four wireless sensor network techniques (E-FEERP, GTEB, HHO-UCRA, EEIMWSN) in maintaining high energy levels. E-FEERP consistently exhibits superior energy efficiency (93 to 98%), emphasizing its effectiveness. The Energy Consumption Dataset provides insights into the joule measurements of energy consumption for each technique, highlighting their diverse energy efficiency characteristics. Latency measurements are presented for four techniques within a fixed transmission range of 5000 m. E-FEERP demonstrates latency ranging from 3.0 to 4.0 s, while multi-hop latency values range from 2.7 to 2.9 s. These values provide valuable insights into the performance characteristics of each technique under specified conditions. The Packet Delivery Ratio (PDR) dataset reveals the consistent performance of the techniques in maintaining successful packet delivery within the specified transmission range. E-FEERP achieves PDR values between 89.5 and 92.3%, demonstrating its reliability. The Packet Received Data further illustrates the efficiency of each technique in receiving transmitted packets. Moreover the network lifetime results show E-FEERP consistently improving from 2550 s to round 925. GTEB and HHO-UCRA exhibit fluctuations around 3100 and 3600 s, indicating variable performance. In contrast, EEIMWSN consistently improves from round 1250 to 4500 s.

Health equity is a fundamental principle that aims to ensure that all individuals, regardless of their background or circumstances, have equal access to quality healthcare. Unfortunately, significant health disparities persist globally, with marginalized and disadvantaged groups often being at a disadvantage in terms of access to diagnostics and treatments. Artificial intelligence (AI) is emerging as a powerful tool to combat these inequalities and improve health equity.

The rise of cyberattacks has necessitated the development of effective malware detection mechanisms. Deep learning, with its ability to learn complex features from raw data, has been widely used for this purpose. This paper presents a two-fold contribution to the field of malware detection using deep learning. Firstly, we use seven pretrained CNN models to classify malware images from the Malimg dataset. While these models achieved high accuracy, they presents low precision and F1-score, indicating that they were prone to false positives and false negatives. Additionally, these pre-trained models are susceptible to overfitting, which is a common issue with transfer learning. To overcome this limitation, we propose a custom CNN model consisting of six layers, trained from scratch on the Malimg dataset. To address potential issues like over-fitting and data imbalance when training models from scratch, we used regularization techniques such as L1 or L2 regularization, dropout. Our proposed custom CNN model outperformed the pretrained models, achieving best/average accuracy values over 5 trials of 100%/98.26%. We also found that our custom model was less susceptible to over-fitting and adversarial attacks. Our proposed approach provides a promising solution to the problem of limitations in using pre-trained models for malware detection.

The advancement of wireless communication and networking technologies has simplified the design of IoT systems. Wireless sensor networks are made up of groupings of sensors at remote locations that cause problems when their batteries die. Energy-efficient pathways must be devised to extend the network’s lifetime. The Self Adapting Low Energy Adaptive Clustering Hierarchy (SA-LEACH) algorithm is offered as a solution, which employs clustering to balance network load and reduce energy usage. This algorithm takes into account parameters such as energy levels and the distance between sensors and a base station. When compared to current methods, the simulation results reveal that SA-LEACH enhances system dependability, longevity, and load balancing.

Technologies dedicated radio infrastructure deployments are no longer adequate to ensure large-scale, low-cost, and reliable communications with the growing adoption of scattered wireless technologies for current services. The goal of this article is to allow the deployment of a radio network infrastructure for numerous client applications, hence building the Internet of Things (IoT), IEEE established IEEE 802.15.4e time slotted channel hopping (TSCH), an extremely efficient, reliable, and predictable time-frequency enabled medium access control (MAC) protocol for the industrial sector, on top of the low-power IEEE 802.15.4 radio. A MAC communication schedule may be constructed from an IEEE 802.15.4e TSCH communication schedule. The IEEE 802.15.4e TSCH definition, on the other hand, makes no mention of how such scheduling may be generated, modified, or maintained. It also has no jurisdiction over the unit in charge of these obligations. This means that the standard is missing the required scheduling mechanism. To meet this demand, several communication scheduling approaches have been presented in the literature. We first establish a novel decentralized communication scheduling approach called mobile scheduling updated TSCH (MSU-TSCH), which implies that network traffic can move in any direction rather than from leave nodes to the root, in this study. The MSU-TSCH algorithm tries to plan time slots by selecting the node that is nearest to the user. While selecting a nearby node, it seeks to assign a roughly equal quantity of dedicated time slots to each member node (or member neighbor). Other scheduling algorithms may not be able to construct better schedules with neighbor nodes as a consequence.

A blockchain is a distributed and decentralized database that allows users to securely store and exchange data without requiring the intervention of a third party. Information, called transactions, is collected in blocks, which are then linked together via cryptographic processes in an irreversible way. The registry has served as a historical record of all actions taken by participants in the network since its launch. An increase in the number of transactions, leading to a considerable increase in the size of the primary blockchains, making them more difficult to maintain, perhaps discouraging certain nodes from storing the entire blockchain and therefore weakening decentralization. In this study, we introduce the low-storage node, a new type of node that stores chunks of blocks rather than whole blocks and is encoded with an erasure code. A low-storage node recovers the initial block by downloading and decoding a enough encoded fragments from other nodes in the network. This strategy has the advantage of allowing certain nodes to keep a reduced version of the blockchain while contributing to its decentralization. This simplifies the scaling of blockchains, which is one of the main flaws of the technology. BlockStock is a complete system that allows nodes to rent out its additional storage space to others. The main innovation is the use of blockchain-based smart contracts that enable frequent, automated and secure payments based on proofs of recovery provided by storage servers.

Abstract Information hiding aims to embed a crucial amount of confidential data records into the multimedia, such as text, audio, static and dynamic image, and video. Image‐based information hiding has been a significantly important topic for digital forensics. Here, active image deep steganographic approaches have come forward for hiding data. The least significant bit (LSB) steganography approach is proposed to conceal a secret message into the original image. First, the lightweight stream encryption cryptography encrypts secret information in the cover image to protect embedded information from source to destination. Whereas the encrypted embedded cover information into the carrier of stego‐image with the help of the LSB and then transmit. In the proposed investigational scheme, a convolutional neural net is used. A model is trained to detect and extract patterns of image hidden features, encrypted stego‐image optimization, and classify original and cover images of steganography. Through the experiment result on the forensic image database for mobile steganography of the Center for Statistics and Application in Forensic Evidence, the overall embedded and extracting that the proposed scheme can achieve information hiding as well as revealing with an accuracy rate of 95.1%. The experimental result shows the robustness of the model in terms of efficiency as compared to other state‐of‐the‐art schemes.

In the transmission of medical images, if the image is not processed, it is very likely to leak data and personal privacy, resulting in unpredictable consequences. Traditional encryption algorithms have limited ability to deal with complex data. The chaotic system is characterized by randomness and ergodicity, which has advantages over traditional encryption algorithms in image encryption processing. A novel V-net convolutional neural network (CNN) based on four-dimensional hyperchaotic system for medical image encryption is presented in this study. Firstly, the plaintext medical images are processed into 4D hyperchaotic sequence images, including image segmentation, chaotic system processing, and pseudorandom sequence generation. Then, V-net CNN is used to train chaotic sequences to eliminate the periodicity of chaotic sequences. Finally, the chaotic sequence image is diffused to change the raw image pixel to realize the encryption processing. Simulation test analysis demonstrates that the proposed algorithm has better effect, robustness, and plaintext sensitivity.

In this paper, a graph-based pan-sharpening technique is proposed which is processed on multiple regions defined as a graph-based superpixels generated from texture descriptor of PAN image. To this end, the PAN texture is to over-segmented into a set of superpixels, which are considered as initialization for the Region Adjacency Graph (RAG). Then, the fusion process is accomplished by inferring details into their corresponding up-sampled MS at feature level guided by the graph-based textural segmentation map. Our proposal has been evaluated using two datasets acquired by the WorldView-2 and WorldView-4 sensors. Its performance is clearly demonstrated in comparison with several state-of-the-art pansharpening methods both at pixel and region levels.

Classification of EEG signals based on motor imagery is an important task in Brain-Computer Interface (BCI). Deep learning approaches have been successfully used in several recent applications to learn features and classify different types of data. However, the number of researches using these approaches in BCI applications is very limited. In this paper, we aim at using the fusion of Convolutional Neural Networks (CNN) methods to improve the classification performance of EEG motor imagery signals in the framework of e-health Internet of Things. We propose and compare two classification methods based on the fusion of two CNNs. Our results show that the fusion of the CNNs with the Long Short-Term Memory (LSTM) layers offers a better classification performance compared to other state-of-the-art methods. The classification performance achieved by our proposed method using the BCI competition IV 2a dataset in terms of accuracy value is 61.68%. This method can be successfully applied to BCI systems where the amount of data is large due to daily recording.

Brain-Computer Interfaces (BCI) based on Motor Imagery (MI) extract commands in real time and can be used to control a cursor, wheelchair, robot or prosthesis by performing only mental imaging tasks, such as imagining a movement of the right hand while the corresponding brain activity is measured and processed by the system. Because MI-based BCI offers a high degree of freedom, it helps people with motor disabilities communicate with the device by performing a sequence of MI tasks. Several techniques are being developed to improve the classification performance of the MI signals used in BCI. Most researches focused on improving methods for feature extraction and selection, but relied on linear classifiers for class prediction. In this paper, we investigate the use of ensemble learning methods to improve classification accuracy in a BCI paradigm based on 2-class MIs. We propose and compare eight combinations of classifiers on the BCI Competition III dataset IVb. The results obtained show that the combination of three classifiers: Radial Basis Function-Kernel Support Vector Machine (RBF-Kernel SVM), Linear Support Vector Machine (Linear SVM) and Decision Tree, gives the best value of kappa which is equal to 0.783. This combination can be successfully applied to BCI systems where the amount of data stems largely from daily recording.

In this work, we propose applying a non-linear strategy to estimate the novel synthetic intensity component through a second-order polynomial regression analysis, such that it could be expressed as a linear combination of multispectral bands in addition to their weighted joint products. Details maps are, therefore, obtained as the residual between the panchromatic (PAN) and the novel synthetic intensity component. Indeed, the high-resolution Multispectral (MS) image is produced by transferring the details into the expanded multispectral bands. The quality improvements achievable by our proposal are evaluated by using two high resolution data sets collected by the WorldView-3 and WorldView-4 sensors. The performance of our proposed technique is clearly shown against the existing state-of-the-art pansharpening methods (especially with linear schemes) and particularly for multispectral channels with non-overlapping spectral wavelengths with respect to the PAN image.

Our paper proposes a new device authentication scheme for mobile sensor node called Mobile Data Collector (MDC). Moreover, to validate the data brought by the MDC to the base station (BS), we validate it and then store it. To solve MDC authentication between multiple devices, we proposed blockchain scheme to provide more ease, communication and security between different devices. For this to happen, the last MDC authentication (meaning the first time the information is gathered) is performed by the CH'S first encounter with the classic authentication, and here the protocol accepts or rejects the MDC. Once the CH has authenticated the MDC, CH sends a transaction to the blockchain to verify the legality of the MDC access. Then, when the MDC requests the collected data from another CH in the network, at this point, any CH verifies the trust of the MDC by communicating with the blockchain. Hence, the proposed scheme is as safe as we claim. More specifically, in the proposed protocol for Blockchain Security IoT (Block_MDC) is to provide authentication between the Mobile Data Set (MDC), the head of the group and the member nodes of the WSN. We evaluate the performance of our protocol using simulations using MATLAB. The results confirm that the Block_MDC protocol is robust, efficient, and offers lower power consumption and fast computing time.

With the emergence of cloud robotics, computation offloading presents a new trend in cloud computing that has been applied to robots; to provide them with resources for performing computationally intensive tasks. In most scientific research, the main objectives behind computation offloading are reducing energy consumption and minimizing the execution time of robotics applications. However, these two metrics are conflicting, and optimizing them simultaneously is challenging. Reducing energy consumption may lead to a rise in the completion time, and vice-versa. In this paper, we consider the problem of optimization of energy consumption and completion time in a cloud robotic system. We formulated the offloading decision as a multi-objective optimization problem. We further adapted the Non-dominated Sorting Genetic Algorithm (NSGA-II) to find a set of Paretooptimal solutions. Through simulations, we demonstrated that our offloading solution can save 80% of the robot’s energy consumption; and reduce 70% of the application completion time. We proved also the adaptability of the model against bandwidth changes.

Pansharpening is achieved by inferring spatial details derived from a PANchromatic (PAN) image into its corresponding expanded multispectral (MS) bands. In this letter, we propose to apply an adaptive superpixel-based injection scheme that modulates the PAN details through an optimization procedure. Optimal injection coefficients can be locally estimated by using the shuffled complex evolution developed in the University of Arizona (SCE-UA) algorithm over multiple local segments (i.e., superpixels) resulting from the simple linear iterative clustering (SLIC) method. The performance of the proposed approach is assessed using degraded and real data sets acquired from WorldView-3 and WorldView-4 satellites. Experimental results show the suitability of the proposed adaptive injection scheme compared with other state-of-the-art pansharpening methods in terms of spatial and spectral qualities.

In response to the researchers need in the bio-medical domain, we opted for automating the bibliographic research stage. In this context, several classification models of supervised machine learning are used. Namely the SVM, Random Forest, Decision Tree, KNN, and Gradient Boosting. In this paper, we conduct a comparative study between experimental results of full article classification and abstract classification approaches. Furthermore, we evaluate our results by using evaluation metrics such as accuracy, precision, recall and F1-score. We observe that the abstract approach outperforms the full article approach in terms of learning time and efficiency.

Cloud robotics is incessantly gaining ground, especially with the rapid expansion of wireless networks and Internet resources. In particular, computation offloading is emerging as a new trend, enabling robots with more powerful computation resources. It helps them to overcome the hardware and software limitations by leveraging parallel computing capabilities and the availability of large amounts of resources in the cloud. However, the performance gain of computation offloading in cloud robotics is still an ongoing research problem because of the conflicting factors that affect the performance. In this paper, we investigate this issue and we design a distributed cloud robotic architecture for computation offloading based on Kafka middleware as messaging broker. We experimentally validated our solution and tested its performance using image processing algorithms. Experimental results show a significant reduction in robot CPU load, as expected, with an increase in robot communication delays.

Nowadays, motorized vehicles are essential in our daily lives, therefore fuel supply services should be efficient and easily accessible. The fuel supply may encounter some difficulties, such as queues, non-continuous availability of fuel, and fraud. The problems of long waiting queues and non-continuous availability of fuel can be solved by going to the next fuel station. However, fraud is a more serious issue that is more difficult to solve. We intend to develop an intelligent system for fuel supply management to solve this problem. For safety reasons, we must avoid the risk of causing a spark in the fueling environment. An electric system close to the pump, the hose, or the vehicle fuel tank may be a risk. We opted for the RFID (Radio Frequency IDentification) technology and the use of passive tags, since semi-passive or active tags involve a battery, on one hand, and are significantly more expensive, on the other hand. A motorized vehicle is identified by a passive RFID tag, and two other passive RFID tags are used for the fueling of all cars from the given fuel pump. Our work consists of the design, by research, of the required system and focuses on the optimization of the topology of antennas and tags so that frauds are prevented. The technique is based on the alignment of tags.

This paper presents a robust object tracking method based on the methodologies of statistical texture analysis of 2D images based on the theory of monogenic signal analysis, jointed with the color histogram. This novel feature extraction method is embedded thereafter in the mean shift framework. Compared with methods of state-of-the-art mean shift trackers, this method proves to be more discriminant and less sensitive to noise. The experimental results proved that our proposed method can achieve robust tracking performances in complex situations with fewer mean shift iterations.

ABSTRACT In this paper, we present a unified algorithm for phase and magnitude speech spectra data hiding. The phase and the magnitude speech spectra are concurrently investigated to increase the capacity and the security of the embedded information. The proposed algorithm in this paper is based on finding secure spectral embedding areas in wideband magnitude speech spectrum. Our approach exploits these areas to hide data in both speech components (i.e., phase and magnitude). The embedding locations and hiding capacity are defined according to a controlled acceptable distortion in the magnitude spectrum. The latter is expressed as a set of parameters controlled by the sender. Consequently, the hiding capacity and the locations of concealed data change for each data communication instance to further prevent malicious intrusions. Objective results show that the presented algorithm in this paper secures hidden data and achieves interesting tradeoffs between the hiding capacity and the speech quality. Copyright © 2013 John Wiley & Sons, Ltd.

In this paper, a robust perceptual audio hashing system is presented. A model of the human auditory system is used to extract robust features from the outputs of a non-linear filter bank that mimics the human basilar membrane. Experiments on various audio excerpts show that this new ear-based front-end processing provides very effective hash values. The proposed audio hashing system performs very satisfactorily in identification and it turned out very resilient to a large variety of severe audio attacks.

Radio Frequency Identification (RFID) has become a popular identification technology in a number of application areas such as supply chain management. A dedicated middleware solution is required to achieve the maximum benefits of RFID technology. The middleware components serve to abstract the communication between the middleware and the different types of sensing devices on one hand, and the middleware and backend applications on the other hand. FlexRFID is a simple and smart RFID middleware which provides device management and monitoring, data processing, filtering, and aggregation, rapid application development, as well as a policy based business rules layer which helps applying rules for accessing and configuring the services provided by the middleware. This layer serves to process some business intelligence rules locally so that the host system is offloaded from those mundane tasks. The paper shows that FlexRFID is a highly scalable and easily deployable middleware in the heterogeneous sites based on different standards and consisting of different hardware. Apart from these, FlexRFID incorporates the mechanisms for supervision, testing, and control of its components, plus handles the security and privacy issues that inhibit the adoption of RFID technology by applying the privacy policies in the business rules layer. FlexRFID middleware is controlling the data flows in and out as well as locally controlling some intelligence.

Ad hoc wireless networks can be described as dynamic multi-hop wireless networks with mobile nodes. However, the mobility condition can be relaxed, and we can consider an ad hoc wireless network as a reconfigurable network where all the nodes are connected to the local environment through wireless links, and where there is not a central or dominant nodes opposed to, for example, the case of cellular wireless networks where a base station is located in each cell. When ad-hoc networks are backboned by fibers, distortion of the optical link presents one of the major issues. In this paper, we will be addressing one of the fundamental problems, namely chromatic dispersion compensation in the fiber optic prior reaching the access points. This will ensure an adequate better quality of signal in the applied network.