Hakim Sabzevari University

The Message Queuing Telemetry Transport (MQTT) protocol is widely adopted in IoT networks due to its lightweight and scalable design, making it ideal for resource-constrained devices. However, its minimal architecture lacks built-in encryption and robust identity authentication, leaving it vulnerable to a range of security threats. To address these vulnerabilities, this work proposes a hybrid framework that integrates machine learning (ML) and deep learning (DL) models to enhance intrusion and anomaly detection in MQTT based IoT networks. The framework leverages real world IoT traffic data from the MQTTEEB-D dataset with diverse preprocessing techniques to train the model. The framework is implemented and evaluated through a multi-metric approach, where each metric assesses a distinct aspect of the framework to identify lightweight processing models for resource-constrained environments. Results consistently showed that ML models outperformed their DL counterparts in terms of detection reliability, classification balance, and error minimization. While DL models demonstrated moderate effectiveness in capturing temporal patterns, they exhibited higher misclassification rates and reduced calibration. The findings underscore the effectiveness of lightweight ML models for scalable and dependable intrusion detection in MQTT based IoT networks.

This study presents a CNN-based deep learning framework to automate the accurate authentication of mericarp seeds from 15 commercially important medicinal Apiaceae species, which exhibit subtle morphological overlaps and face risks of toxic adulteration (e.g., hemlock contamination), threatening consumer safety and trade integrity. Six CNN models were evaluated, with DenseNet121 demonstrating superior performance in terms of accuracy, precision, recall, F1-score, and convergence stability, followed by MobileNetV2, InceptionV3, and VGG16. Confusion matrix analysis revealed that higher-resolution inputs significantly improved discrimination, particularly for similar seeds like Anethum graveolens L. and Apium graveolens L. Batch size minimally influenced outcomes. Analysis of accuracy-loss dynamics further indicated that EfficientNetB0 and ResNet50 underperformed, while DenseNet121, which excelled in performance, convergence stability, strong generalization, and minimal overfitting, highlighted the critical role of architectural design in feature learning and optimization of key performance metrics. Additionally, t-SNE visualization confirmed DenseNet121's superior feature learning, achieving distinct separation of all 15 species and resolving intricate visual ambiguities that confound traditional methods and limit clearer clustering in other architectures. These findings underscore the potential of CNNs as scalable tools for botanical authentication-particularly for detecting adulteration and species complexities using seed digital morphometric characters-to safeguard public health and reinforce supply-chain safety.

Image enhancement remains a cornerstone in digital image processing, aiming to improve visual clarity through various methods. Spatial domain techniques include integer-order and fractional-order differentiation. Although widely used, traditional integer-order differentiation techniques suffer from limitations such as indiscriminate spatial frequency treatment and noise amplification, leading to degraded image quality. This paper proposes an adaptive fractional-order differentiation approach employing Caputo fractional differential masks to selectively enhance image details. This approach uses image gradient information to determine the appropriate fractional order. By dynamically adjusting the fractional order based on specific image requirements, the method achieves superior contrast improvement while preserving fine details and minimizing noise. Experimental results, evaluated using metrics such as Pratt's Figure of Merit (FOM), Structural Similarity Index (SSIM), and Peak Signal-to-Noise Ratio (PSNR), demonstrate that this approach outperforms comparable techniques, highlighting its effectiveness in image enhancement.

The present study investigates the dynamics of land-use changes and their relationship with dust generation in the semi-arid region of Fars Province. The main objective of this study is to understand how the interactions between land-use patterns and dust-generation intensity affect the environment. Using remote-sensing data and advanced modeling techniques such as CA-Markov and meta-heuristic algorithms, land-use changes and their impact on dust occurrence are predicted for the years 2024, 2050, and 2100. The results show that land-use changes, especially the reduction in agricultural land and increase in bare land, play an important role in the intensification of dust generation. Predictions indicate that by 2100, agricultural land will have decreased significantly, and bare land will have increased sharply. These changes, which are primarily caused by anthropogenic processes and urban development, will lead to the expansion of high-risk dust zones in the area. Modelling shows that Ant colony optimization ( ACO) is superior to other methods in making proper predictions of these changes, as well as assessing dust risk. The study reveals the relevance of future land-use change prediction to manage the vulnerability of dust sources in Iran's arid and semi-arid regions. The study reveals that future agricultural changes, urbanization, and infrastructure expansion, will directly and prominently influence the intensity and pattern of dust. Understanding such trends can help policymakers carry out adequate management policies that will reduce the negative impact of dust on ecosystem health and air quality. The results of this research, in addition to advancing scientific knowledge of environmental processes, hold out the possibility of developing long-term policies that will ensure public health and natural resource conservation in vulnerable regions.