Robust and intelligent intrusion detection is vital for securing Internet of Things (IoT) ecosystems against evolving cyber threats. However, existing systems face challenges such as class imbalance, suboptimal model performance due to manual hyperparameter tuning, and the high cost of labeled data. These limitations are addressed using the TON IoT dataset. To resolve data imbalance, the proximity weighted random affine shadow sampling generates boundary-focused synthetic samples that preserve class distribution. Further, to tackle suboptimal performance, bayesian optimization is applied to LeNet, resulting in LeBayesNet, which discovers the optimal configuration for high-accuracy detection. Next, to mitigate the scarcity of labeled data, MargiLeNet leverages marginal-based active learning, annotating the most uncertain samples to enhance model learning efficiently. Experimental results show that LeBayesNet and MargiLeNet improve performance over existing models by 7.69% and 3.30% in accuracy, 7.69% and 3.30% in F1-score, 7.69% and 3.30% in precision, 8.89% and 4.44% in the recall, 7.69% and 6.59% in receiver operating characteristic-area under the curve, 4.88% and 7.32% in matthews correlation coefficient, and 10.34% and 11.49% in precision-recall area under the curve, respectively. Both models significantly reduce hamming loss to 75% and 37.5%, indicating better generalization in complex and imbalanced scenarios. These advancements demonstrate the potential of optimization and active learning techniques in building accurate and adaptive intrusion detection systems for modern IoT networks.

This paper proposes a DL based framework using Multilayer Perceptron (MLP) tailored for multiclass DoS attack detection in Internet of Things (IoTs). After comparative data preprocessing, class imbalance is effectively mitigated using the proximity weighted random affine shadow oversampling method, enhancing minority class representation. Moreover, feature selection based on variance threshold is employed to streamline the input space and accelerate training. To reduce dependence on large labeled datasets, the approach incorporates Diversity-Based Sampling (DBS), an active learning strategy that focuses labeling efforts on diverse, informative samples. Furthermore, the proposed model’s performance is refined through metaheuristic-driven hyperparameter tuning using the Grasshopper Optimization Algorithm (GOA). This integrated methodology ensures more efficient learning, better generalization, and improved detection across varied attack scenarios in IoT settings. A comparative analysis with traditional machine deep learning and baseline models reveals that the proposed MLP+DBS and MLP+DBS+GOA model configurations consistently deliver superior performance across all evaluation metrics. Specifically, the proposed models achieve improvements of 5.7% and 9% in accuracy, 3.5% and 8.3% in precision, 5.7% and 9% in recall, 3.4% and 8.6% in F1-score, 2.3% and 3.3% in receiver operating characteristic area under the curve, and 3.3% and 6.6% in precision recall-area under the curve, respectively. These results demonstrate that the proposed models significantly outperform the existing approaches. This paper underscores the effectiveness of combining active learning and optimization for robust intrusion detection in resource-constrained IoT settings. The proposed models show strong potential for real-time deployment in smart environments requiring proactive and reliable security solutions.

With the rapid increase of connected devices, securing IoT networks against sophisticated cyber threats has become a critical research priority. However, effective intrusion detection in IoT environments is hindered by several core challenges, including severe class imbalance in network traffic, limited availability of annotated data for supervised learning, and the sensitivity of deep learning models to hyperparameter configurations. To address these limitations, we propose a data-driven DL framework that combines data balancing, active learning, and hyperparameter optimization. We employ the proximity-weighted synthetic oversampling technique to mitigate class imbalance by generating weighted synthetic samples. To reduce labeling overhead, we propose an active learning-based, Entropy-based Convolutional Neural Network (EntroConvNet), an intrusion detector for selective annotation of the most uncertain samples. Additionally, a novel Random Search Optimized Convolutional Neural Network (RS-ConvNet) is proposed to maximize detection performance. Experimental results on the TON IoT dataset show that EntroConvNet outperforms the baseline models with improvements of 3.45% in accuracy, 3.57% in precision, 1.10% in recall, 2.30% in F1-score, 3.19% in Area Under the Receiver Operating Characteristics Curve (AUC-ROC), 6.67% in Cohen’s Kappa and Mathews Correlation Coefficient (MCC), and 16.67% reduction in log loss and Hamming loss. Furthermore, RS-ConvNet achieves superior gains of 4.60% in accuracy, 5.95% in precision, 1.10% in recall, 3.45% in F1-score, 2.13% in AUC-ROC, 8% in Cohen’s Kappa and MCC, and also reduces log loss and Hamming loss by 20% and 25%, respectively. These results validate the proposed framework’s ability to deliver accurate, and annotation-efficient intrusion detection systems in dynamic IoT network environments.

The rapid expansion of Internet of Things (IoT) networks has made them increasingly vulnerable to diverse cyber threats, necessitating the development of efficient Intrusion Detection Systems (IDS). Traditional models for IDS often face challenges such as data imbalance, scarcity of labeled samples, and suboptimal performance due to manual hyperparameter tuning. To address these issues, we propose a comprehensive IDS framework comprising three key components. First, we mitigate data imbalance using the proximity weighted synthetic oversampling technique, which enhances class distribution, followed by the use of Pointer Network (PtrNet) for classification due to its ability to model variable-length sequential data. Second, to handle the scarcity of labeled data, we introduce an entropy-based active learning strategy on PtrNet, termed Entropy-based Active Learning Pointer Network (EAL-PNet). Finally, we optimize model performance through harris hawk optimization applied to PtrNet, resulting in Hawk-Pointer Attention Network (HPA-Net). Experimental results demonstrate that the proposed models significantly outperform traditional approaches. EAL-PNet achieves a performance improvement of 9.30% in accuracy, 8.14% in F1-score, 8.14% in precision, 9.30% in recall, 3.16% in Receiver Operating Characteristic - Area Under the Curve (ROC-AUC), 13.92% in Matthews Correlation Coefficient (MCC) and Cohen's Kappa, and 45.71% reduction in log loss. Similarly, HPA-Net shows a 10.47% gain in accuracy, 10.47% in F1-score, 9.30% in precision, 10.47% in recall, 2.11% in ROC-AUC, 15.19% in MCC and Cohen's Kappa, and 51.43% decrease in log loss. These findings validate the effectiveness of the proposed framework in enhancing intrusion detection for IoT environments.

The increasing complexity and interconnectivity of Internet of Things (IoT) ecosystems have heightened the need for robust and intelligent intrusion detection mechanisms. However, the development of effective detection models is impeded by challenges such as imbalanced data distributions, limited availability of labeled samples, and the difficulty of tuning deep learning architectures to accommodate diverse threat patterns. In response to these challenges, this paper introduces two novel DenseNet-based frameworks, DN-UBS and DN-GBO, for advanced IoT intrusion detection. The proposed approach begins by applying a variance threshold technique on RT-IoT2022 dataset, to eliminate low-variance features, followed by synthetic minority oversampling technique to alleviate class imbalance and enhance minority class representation. DN-UBS integrates an uncertainty-based sampling strategy to iteratively select the most ambiguous instances for annotation, reducing labeling effort while improving model discriminability. In contrast, DN-GBO incorporates a gradient-based hyperparameter optimization using the hyperband strategy, allowing for automatic adjustment of network depth, learning rate, and regularization parameters. The DN-GBO achieved superior detection performance with an improvement of 7% in accuracy, 4% in F1-score, 12.8% in precision, 6 % in recall, 3 % in Receiver Operating CharacteristicArea Under the Curve (ROC-AUC), and 17.6 % in Matthews Correlation Coefficient (MCC). Similarly, DN-UBS also delivered high efficacy with an improvement of 5.3 % in accuracy, 3 % in F1-score, 3% in precision, 4.6% in recall, 4% in ROC-AUC, and 10.1 % in MCC, while minimizing reliance on labeled data. These findings highlight the effectiveness of the proposed models in delivering scalable, adaptive, and data-efficient solutions for securing IoT infrastructures against intrusive threats.

Blockchain-based smart contracts, while transformative, pose privacy concerns due to Ethereum's transparency. To address this, we present Safe Smart Contracts (SafeSC), leveraging zk-SNARKs for privacy without compromising Ethereum's transparency. SafeSC's Python tool facilitates contract understanding and verification without accessing the source code. Our paper explores privacy preservation techniques, favoring Zero-Knowledge Proofs (ZKPs). SafeSC employs zk-SNARKs and Groth-16, achieving a delicate balance between transparency and privacy in smart contract development. The tool’s design, covering architecture, assumptions, data flow, and zero-knowledge proof workflow, marks a step toward secure smart contract solutions. We advocate for continued exploration and refinement to enhance blockchain technologies.

In recent years, Long language models have made significant progress, enabling machines to interpret and process human language. However, the Arabic language presents unique challenges due to its rich morphology and diverse sentence structures. The development of specialized language models for Arabic question answering has implications for improved human- computer interaction, cultural preservation, and accessibility. This paper aims to enhance the comprehension and contextual understanding of Arabic-posed questions by leveraging the capabilities of the LLaMa language model and the XLNet transformer. The ARCD dataset, which mainly consists of an Arabic dataset for question-answering, was used to fine- tune the LLaMa 2.0 and XLNet. By utilizing LLaMA and XLNet transformers separately, This paper contributes to the construction of an NLP pipeline that can properly understand and process Arabic text to provide answers depending on a particular Arabic context by using LLaMA and XLNet transformers individually. It is important to note that Arabic datasets were not previously used to train the LLaMa language model. The LLaMA language model received accuracy scores of 93.70

The perfume industry’s complex and multi-faceted supply chain is challenged by issues of transparency, accountability, and efficiency. To address these issues, this paper proposes a ScentTrack blockchain-based solution for the perfume manufactory, that provides benefits such as improved product quality, supply chain efficiency, and increased trust among supply chain partners. The system was created in collaboration with a perfume manufacturer based in Saudi Arabia, and it incorporates multiple chaincodes distributed across multiple channels. These chaincodes serve to monitor and track the various stages of the perfume supply chain. The proposed solution leverages the automation capabilities of blockchain to enhance trust and accountability among participants. To facilitate this, each participant will have access to their own application portal, enabling them to manage and monitor their respective supply chain. Moreover, the system includes extensive event notifications that keep all stakeholders informed about the multiple steps within each stage of the supply chain they are interested in.Based on the findings, the study suggests that a blockchain system for perfume manufacturing supply chains could transform the industry by establishing a more efficient, secure, and transparent network, which lowerultimately lowering costs while improving quality. Additionally, the proposed solution could be adapted for other supply chains, although further research is required to assess its applicability in different industries. The implementation of this blockchain-based solution in the perfume industry could offer a practical demonstration of how blockchain technology can address the challenges of supply chain management.

Agriculture is an important sector that plays an essential role in the economic development of a country. Each year farmers face numerous challenges in producing good quality crops. One of the major reasons behind the failure of the harvest is the use of unscientific agricultural practices. Moreover, every year enormous crop loss is encountered either by pests, specific diseases, or natural disasters. It raises a strong concern to employ sustainable advanced technologies to address agriculture-related issues. In this article, a sustainable real-time crop disease detection and prevention system, called CROPCARE, is proposed. The system integrates mobile vision, Internet of Things (IoT), and Google Cloud services for sustainable growth of crops. The primary function of the proposed intelligent system is to detect crop diseases through the CROPCARE—mobile application. It uses the superresolution convolution network (SRCNN) and the pretrained model MobileNet-V2 to generate a decision model trained over various diseases. To maintain sustainability, the mobile app is integrated with IoT sensors and Google Cloud services. The proposed system also provides recommendations that help farmers know about current soil conditions, weather conditions, disease prevention methods, etc. It supports both Hindi and English dictionaries for the convenience of the farmers. The proposed approach is validated by using the PlantVillage data set. The obtained results confirm the performance strength of the proposed system.

The recent COVID-19 (novel coronavirus disease) pandemic induced a deep polarization among regional as well as global communities. The sentiments regarding the pandemic and its impact on lifestyle and economy, often expressed via social networks, are regarded as critical metrics for capturing such polarization and formulating appropriate intervention by the relevant authorities. While there exist a myriad of Natural Language Processing (NLP) models for mining social media data, we demonstrate the shortcomings of the individual models in this paper, and explore how to improve the COVID-19 sentiment analysis in social media network data via two hybrid predictive models based on a Long-Short-Term-Memory (LSTM)-based autoencoder and a Convolutional Neural Network (CNN) model coupled with a bi-directional LSTM. Through extensive experiments on the recently acquired Twitter dataset, we compare the COVID-19 sentiments exhibited in the USA and Canada using our proposed hybrid predictive models and demonstrate their superiority over individual Artificial Intelligence (AI) models.

As a consequence of the global pandemic, many restrictions and rules were enforced. One predicament was the travel restrictions and requirements put into place with regard to vaccinations. Countries worldwide now require people to be vaccinated upon entry. The process of validating vaccine doses requires lots of paperwork and is inefficient. Blockchain is an uprising technology that is secure and fast at carrying out transactions. We propose implementing vaccine dose verifications between countries through vaccine certificates using Blockchain as an effective solution. The need for a common shared database, avoiding a trusted third party to administrate the network, having several countries involved, ensuring privacy and security, and accountability logs make Blockchain needed in this scenario. Digital vaccine certificates are very sensitive information that must be kept private and secure but accessible to several entities. Blockchain ensures the aforementioned requirements are met while preserving the integrity of the VDCs. This paper describes blockchain technology and its application in digital vaccine certificates.

Stability of the ideal plant environment in a greenhouse can be maintained by using wireless sensor networks, which are used for monitoring and controlling temperature, light, and humidity. Tracking plant growth is the best method for early detection of disease thus preventing significant crop losses. Wireless Visual Sensor Network (WVSN) are used for monitoring plant growth with the added feature of a camera. This paper presents a mathematical formulation and an optimal solution for the placement of the WVSN cameras to guarantee coverage of a large area while maintaining high quality images and minimizing overlap between cameras. Simulation results show the effectiveness of the proposed model in finding the minimum number of cameras with the exact position to cover the entire monitored area of the greenhouse, with the desired image quality resolution.

COVID-19 (Coronavirus) is a very contagious infection that has drawn the world public’s attention. Modeling such diseases can be extremely valuable in predicting their effects. Although classic statistical modeling may provide adequate models, it may also fail to understand the data's intricacy. An automatic COVID-19 detection system based on computed tomography (CT) scan or X-ray images is effective, but a robust system's design is a challenging problem. In this paper, motivated by the outstanding performance of deep learning (DL) in many solutions, we used DL based approach for computer-aided design (CAD) of the COVID-19 detection system. For this purpose, we used a state-of-the-art classification algorithm based on DL, i.e., ResNet50, to detect and classify whether the patients are normal or infected by COVID-19. We validate the proposed system's robustness and effectiveness by using two benchmark publicly available datasets (Covid-Chestxray-Dataset and Chex-Pert Dataset). The proposed system was trained on the collection of images from 80% of the datasets and tested with 20% of the data. Cross-validation is performed using a 10-fold cross-validation technique for performance evaluation. The results indicate that the proposed system gives an accuracy of 98.6%, a sensitivity of 97.3%, a specificity of 98.2%, and an F1-score of 97.87%. Results clearly show that the accuracy, specificity, sensitivity, and F1-score of our proposed system are high, and it performs better than the existing state-of-the-art systems. The proposed system based on DL will be helpful in medical diagnosis research and health care systems.

Monitoring the climate is one of the most important and challenging practices by which to obtain optimum crop production in a greenhouse. In a smart greenhouse, a wireless sensor network (WSN) can be used to monitor the microclimate. Constant monitoring and sensing can result in excessive energy consumption. Prediction of the microclimate can be used to control the operation of sensors and hence lower the energy consumed by sensor nodes. We develop a Long Short-Term Memory (LSTM) based on time series for the prediction of the maximum, minimum, and mean values of the air temperature, relative humidity, pressure, wind, and dew point. Microclimate data inside and Macroclimate data outside the greenhouse are collected daily and used for the analysis of the best-fitting LSTM model. After determining the network structure and parameters, the network is then trained. The statistical criteria for measuring the network performance are the Mean Absolute Error (MAE), Mean Square Error (MSE) and Root Mean Square Error (RMSE). A comparison is made between the measured and predicted values of temperature, relative humidity, pressure, dew point and wind. Results indicate the effectiveness of the predictive model performance LSTM in predicting the microclimate. Statistical analysis of the RMSE and MAE results demonstrate the prediction accuracy of our proposed LSTM model.

The problem of managing congestion and overcrowding in a critical situation has largely been studied over the last decades. The problem is how to safely direct pedestrians with suitable velocity to the nearest floor in order to avoid the bottleneck. For example, in Hajj there is a ritual practice where millions of pilgrims arrive in a certain area and we should indicate the nearest and lowest congestion path that leads to the ritual. In this paper, we address this problem during the Hajj season and propose a crowd management service based on the game theory model. We modeled the problem as a Mean-Field-Game (MFG) where the solution is a system composed of a Backward Hamilton-Jacobi-Bellman equation and a Forward Transport (Kolmogorov) equation. Simulation results show the efficiency of the MFG solution on the total time out of the pilgrims to reach the ritual.

Sensors, RFID, Wi-Fi, and other technologies embedded with devices and items such as home appliances, vehicles, and grocery items improve the quality of life by exchanging information among each other under a common network platform that defines the emerging future of the Internet, also known as Internet of Things (IoT). Sensors or Wireless Sensor Networks (WSNs) consist of an integral part of IoT since sensors can be controlled by end users and data can be transmitted to distant sites through Internet. Moreover, thousands of sensors and similar devices poses a great challenge in routing that raises the need for zone-based (or cluster) based routing protocol. Most existing routing protocols are not designed considering the dense architecture of IoT. It is a great challenge to render these algorithms adaptive to the changing requirements of sensor-based IoT applications since their routing policies are mostly predetermined. Thus, they are not energy efficient and fault tolerant for such mobility centric IoT. In this article, we provide a brief introduction to IoT with the current state-of- the-art research and classify routing protocols based on several factors. We then introduce a Multiple Base station and Packet Priority-based Clustering scheme (MBPP) for IoT and evaluate its performance through simulations.

Wireless Sensor Networks (WSNs) have been achieved widespread applicability in water quality monitoring. However, existing WSN-based monitoring systems are not adequate for monitoring pond and lake water, city water distribution and water reservoir. Moreover, these frameworks cannot be reused in other monitoring applications since they use static and application specific sensor nodes and are not dynamic to the changing requirements. Thus, we introduce a reusable, self-configurable, and energy efficient WSN-based water quality monitoring system that integrates a Web-based information portal and a sleep scheduling mechanism of sensor nodes. The testbed and simulation results show that the framework can monitor the water quality in real-time and the sleep scheduling mechanism increases the network lifetime, respectively.