Journal of Al-Qadisiyah for Computer Science and Mathematics

This study examines the crucial relationship between Artificial Intelligence (AI) and statistics, which has grown in importance due to the information boom and computational advances. Modern data-intensive applications have led these domains to merge, despite their parallel history. This document explores AI\'s fundamentals, focusing on machine learning and deep learning, and how statistical methods help AI perform tasks like decision-making and pattern identification. We integrate AI with robust statistical modelling and prediction to improve AI transparency and effectiveness. This multidisciplinary approach emphasizes theoretical advances, practical applications, ethical considerations, and future problems at the interface of AI and statistics. We encourage AI and statistics communities to collaborate to promote innovation and responsible AI development and deployment. This collection of publications is a comprehensive resource for researchers and practitioners using AI and statistics to better decision-making and predictive analytics

With the rapid development that the world has witnessed in all scientific fields, data analysis has become more complex due to the large size of collected data that needs to be analyzed. Because of that the data analyst needs a powerful tool to deal with such that kind of data and this is why the dimension reduction methods have developed. In this paper, we focuses on minimum average variance estimation as powerful dimension reduction tools to analysis of ischemic stroke for sample of patients under some covariates. The aim of this paper is to determine the most important covariates that affect the size of ischemic stroke as response variable.
And the results indicated that the SMAVE method is the best method for prediction of the future ischemic stoke volume.

This paper describes a novel integral transform called the "Dukani Transform," which we can employ to precisely (analytically) solve the Volterra integral equations (VIE) of the first sort. To do this, first the Dukani transform of transcendental and elementary algebraic mathematical functions is derived by the authors. Next, the basic properties of the Dukani transform were discussed. Then, using a range of differential equations, including integral equations (IEs), ordinary differential equations (ODEs), and integro differential equations (IDEs), the exact (analytic) solution for first-class generic VIE was found by the authors. Numerical problems have been thoroughly examined and solved, in a step wise manner, demonstrating the applicability of the Dukani transform. The findings demonstrate that, without need for time-consuming calculations, the suggested "Dukani Transform" new integral transform produces accurate (VIEs.) of first-kind and second kind.

In this paper, a new adaptive parametrization model of fourth order neuron cells model of Hodgkin-Huxley (HH) is introduced. The adaptation algorithm design or as can be called adaptive identification algorithm of parameters of the neuron cells HH model is produced. This algorithm is theoretically and computationally proved. This introduced algorithm is basically based on Lyepunov functions and adaptive observer methods which may prove the stability of the model to obtain unique solutions. The results of the computer simulation of the identification problem are shown in the figures and the parameters are clearly coincide the real values. Some terms of the problem are defined for the measured data. It provides an observation model to determine the most informative data for a specific parameter, and find the best fit model. In the HH model of neural cells, the consistency of the differential equations of the adaptation model with the observers of cell’s activation is shown by fitting the observed data to the real data within the high accuracy of estimating the parameters

This research aims to use the linear programming method to identify underutilized production capacities within the company's production lines by employing the parametric programming method as a supporting tool. It is considered a fundamental and important method that helps make the best use of the resources available to companies and aids decision-makers in making accurate, scientifically-based decisions. The study includes identifying unused capacities within four production lines of a dairy company: the dairy production line and the confectionery production line. The study identifies unused capacities within the natural juice and beverages production line and the bakery production line. Based on the results obtained and a comparison with actual achievements, it identifies the unused production capacities. This approach grants the model greater flexibility and uses it as a foundational basis for determining optimal values in the future.

In this article, numerical method based on Laguerre-Krall polynomials for solving nonlinear integral equations (NIE) of the first type which is a malignant problem is discussed. This method reduces the operation of solving the problem and turns it into simple systems of obvious algebraic equations that are easily solvable. Four examples were presented to analyze the numerical method, applicability degree and accuracy of the method. Numerical results showed that the accuracy of this method is acceptable and good compared to other methods.

This study presents a novel subclass of analytic and bi-univalent functions within the open unit disk, utilizing Horadam polynomials related to Gamma-starlike functions. Our primary focus is on deriving upper bounds for the second and third Taylor-Maclaurin coefficients of functions belonging to this subclass. We employ subordination techniques and properties of Horadam polynomials to establish these coefficient estimates.

This paper introduces and defines subclasses of the function class ∑ of analytic and bi-univalent functions associated with the operator N_λ^n (α,β,μ) f(z) within the open unit disk through quasi-subordination. We derive results concerning the corresponding bound estimations of the coefficients a_2 and a_3.

This study aims to demonstrate the process of three different models for predicting the number of undergraduate students enrolled at the Education College of Benghazi in a certain period. These models are Exponential change method, logistic method and the least squares method. Data was collected from registration and admissions office of the education college of Benghazi for the years 2015-2023. This data will be used to forecast for all methods that will be discussed in this research. For evaluating and comparing the efficiency and accuracy of these methods, Mean Absolute Deviation (MAD), Mean Absolute Percent Error (MAPE), and Root Mean Square Error (RMSE) were used. Enrollment data analysis allowed for the comparison of the models to the actual enrollment in regards to which method gave the best prediction of future student enrollment. The least squares method provided the most accurate estimations, with the precision rate of 92%. The three models were further used to predict enrollment levels for the Education College of Benghazi in the future semesters.

The principle idea of our work is to present newly generalization for two major types of modules, which are hollow module and lifting module they are supplement –hollow module and supplement –lifting module respectively. We consider supplement - hollow module as. Let C be a non-zero unital left R-module and R be a ring with identity then C is known as the supplement-hollow module ( indicates sp-hollow). If each proper submodule of C is a supplement-small submodule in C. Furthermore, an R-module C is known as supplement –lifting module (indicates sp-lifting).If there is a submodule H of U for each sub module U of C such as C=H⨁ L and U∩ L ≪_sp C ,where L is a submodule of C. After studying these ideas, we came up with some connected findings

The coronavirus COVID-19 pandemic has caused a global health crisis. According to According to the World Health Assembly, one of the best preventative measures is to wear a face mask while out outdoors (WHO). This work presents a hybrid model for face mask identification that combines deep and traditional machine learning. I have trained the proposed system, which consists of convolutional neural networks (ConNN), support vector machines (SVM), and random forests (RF), in three stages, the first stage, used ConNN, the second stage, used the same ConNN with the SVM method, and in the third stage, used ConNN and RF. This paper suggests three different kinds of masked face recognition datasets: the Incorrectly Masked Face Dataset (IMFD), the Correctly Masked Face Dataset (CMFD), and the combination for MaskedFace-Net, a worldwide masked face detection system. Two objectives are presented for the realistic masked face datasets: i) to identify individuals whose faces are covered or not covered, ii) to identify faces whose masks are put on properly or improperly (for example, at airport entrances or among crowds). The suggested model is made up of two parts. The first part is designed for feature extraction using a convolutional neural networks. In contrast, the second section is made to classify face masks using SVM and RF methods. The ConNN achieved 99.92%. and achieved for ConNN and SVM 99.94%. ConNN and RF 98.79%. Moreover,The system has been tested in real world scenarios and can recognize and classify any image selected by Google with high accuracy. we a comparison and the results aim to evaluate the proposed model.

The Internet of Things (IoT) has brought about a new age in the electric power sector, allowing better processing, safety, and price management. IoT systems work via the supply chain to transport, check, and share the collected data. They get the data that will help to make the right decisions from different sources, and then they further process and interpret it. By this, the decision-making skills are improved. Internet of Things is about collecting, organizing, and exchanging data in different locations and across platforms. It's a technology that shortcuts our metros but by real-iconic monitoring and self-detecting we are sure to receive power on time and when faults occur it addresses them immediately. IoT setup has given rise to Smart Grids, which are hi-tech, self-regulated communication networks employing processes like artificial intelligence and deep learning that enable quality, energy, usage regulation and real-time grid repairs and security.

One technique for creating visuals that correspond to textual descriptions is called "text-to-image generation." It affects a wide range of applications and research fields (e.g., photo-editing, photo-searching, art-making, computer-aided design, image reconstruction, captioning, and portrait drawing). With the development of text-to-image generation models, artificial intelligence (AI) has reached a turning point where robots are now able to convert human language into aesthetically beautiful and coherent images, creating new opportunities for creativity and innovation. The creation of stable diffusion models is one of this field's most noteworthy developments. These models provide a strong framework for producing realistic images that are semantically linked with the given textual descriptions. But even with their remarkable abilities, conventional text-to-image models frequently have serious shortcomings, especially when it comes to training timeframes and computing costs. These models can be costly and time-consuming to train because they usually need large amounts of processing power and long training times. The main goal of this work is to develop a better Stable Diffusion model to overcome these shortcomings and produce high-quality images from text. The suggested model will drastically cut down on training durations and processing needs without sacrificing the quality of the output photos. The proposed method shows that the fine-tuning of the Stable Diffusion model results in a considerable improvement in producing images that are more akin to the original. The results of the improved model denoted a lower FID score (212.52) when contrasted with the base model (251.22).

ICT is becoming more and more popular in the field of remote control. In the healthcare system, patients are monitored in an intensive care unit after a surgical procedure until they are physically stable, then moved to a room for evaluation and recovery. Usually, ward evaluation does not imply continuous monitoring of physiological parameters, and therefore patient relapse is not uncommon. This paper describes the steps taken to design and build a prototype for a low-cost, modular monitoring system. This system is intended to provide mobile support to facilitate faster and better medical interventions in emergencies and has been developed using dedicated low-power sensor arrays for BPM, SpO2, and temperature, as well as room temperature and humidity. The interfaces for these sensors are developed according to the IoT model: the central console displays a web interface based on a REST API that ensures platform-neutral behavior and provides a flexible mechanism for integrating new components. Finally, this paper also investigates the technologies and systems related to e-health services with a better understanding of monitoring applications based on multiple models and different IoT sensors. Finally, this study contributes to scientific knowledge by identifying the main challenges of the topic and providing possible opportunities in this research area.

Organizations and individuals increasingly rely on cloud services for data storage and transmission, and the need for robust and efficient encryption algorithms is more critical than ever. This paper rigorously examines the performance and NIST compliance of two lightweight encryption algorithms, RC6 and PRESENT, within cloud computing contexts. The paper assesses the algorithms' performance metrics, such as encryption time, decryption time, space-time, memory used, CPU consumption, avalanche effect, entropy, and energy consumption, in addition to their security features. Also, utilizing fifteen NIST SP 800-22 cryptographic tests to ensure secure and efficient practices, including recent security analyses to assess the two algorithms' behavior, both algorithms present a distinct array of compromises between effectiveness and protection

In recent years, Machine Learning (ML) has brought about a significant revolution in several fields such as medicine, justice, cybersecurity, and other vital fields that require intelligent and urgent decision-making. With this development, a type of adversarial attack targeting ML models called a Poisoning Attack (PA) has emerged. One realistic attack scenario is for an adversary to subtly update samples or reverse some labels of training data, causing degradation to the model's overall accuracy during the testing phase. To gain a deeper understanding of this scenario, a survey will be conducted about the attack and how it is carried out against different models. In addition to the protection techniques to identify their weaknesses. Finally, some solutions will be proposed to maintain the availability, robustness, and integrity of ML models.

Recently, a greater emphasis in research is on the targeting the parameters responsible for the spread of diseases, particularly in managing the complexities of large datasets related to disease information One of the major problems is trying to attain a high level of precision because some data sets in big data can be incomplete. The scope of this research involves advanced learning systems to formulate a system which provides best course of action against treatment recommends of any disease. Among other approaches, dealing with the missing data points and regularization of disease databases are included. The Whale Optimization Algorithm (WOA) will be developed to enhance predictions of effective treatments for diseases, utilizing genetic algorithms, which have unique features that set them apart from other methods. The results of the proposed approach showed significant improvement in predicting the appropriate treatment for diseases, compared to earlier results obtained with the WOA algorithm before its enhancement. The new method demonstrated higher accuracy reaching 98%.

One application of Natural Language Processing (NLP) is Next Word Prediction, also known as Language Modeling. This process involves predicting the most likely word to follow in a given sentence based on the preceding context. It has numerous widely used applications, like auto-correct, which is mostly used in emails and messages. It can also be used in Microsoft Word or Google searches to predict the next word based on past searches or global queries. The goal of Natural Language Generation (NLG) is to create language that is human-interpretable and natural. Users find text generation, and next-word prediction in particular, convenient as it makes typing faster and error-free. Consequently, an essential analysis topic for all languages is a personalized text prediction system. This paper suggests a novel approach for predicting the following word in a Arabic sentence. It is possible to minimize the total number of keystrokes a user makes by anticipating the next word in a sequence. In this work, Bert algorithm and Continuous Bag of Words(CBOW) are proposed to predict the next word in Arabic language, and predict the previous word. The Bert Algorithm is achieved the best accuracy , 90% for next word prediction, and 80% for previous word prediction . And, Continuous Bag of Words(CBOW) is achieved the best accuracy , 100% for next word prediction, and 100% for previous word prediction.

Dark web emerged as an encrypted and closed network that can only be accessed by computers using specific software and allows users to apply for membership and you have to provide your IDs to the admin of that network, which creates severe challenges in passing cybersecurity; With advanced Internet facilities, the origin and possibility of network invasions or attacks are increasing, and it is becoming very difficult for traditional anomaly detection systems to analyze objectively and pass Functional Purpose: Our ultimate goal by applying machine learning algorithms and network analysis methods is to identify a certain set of attributes that May indicate a cyber threat. Like many other studies, the traffic in the current paper has also been generalized to represent darknet traffic with a classification of unique features between attacks and attacks. As is known, based on the results of the experimental analysis of the proposed approach, it is possible to conclude about the high efficiency of the developed method in identifying types of cyber threats . This project contributes, by utilizing and improving machine learning techniques, to finding reliable measures to mitigate threats in cyberspace, and improving threat detection programs.

Exploring innovative communication technologies has been a pressing need due to the need for reliable and efficient data transmission in networks such as industrial automation, IoT devices, and healthcare systems. This research investigates a wireless communication system that uses visible light, called microcontrollers, with Li-Fi technology instead of radio waves to enhance data accuracy and resilience. Unlike traditional legacy systems, Li-Fi technology provides a buffer against electromagnetic interference, making it a promising alternative for smart home automation. This research includes a developed prototype that uses visible light communication (VLC) through LEDs and an LDR receiver, controlled by an Arduino microcontroller. The KNX standard was adopted in building the proposed system, which has been widely adopted in building automation. The results showed that the Li-Fi-based system improves the accuracy of data transmission as well as the data transmission power. This research, along with other researches on Li-Fi, contributes by demonstrating its applicability in smart home environments, and thus this technology will revolutionize wireless communications across various fields.

Globally, cloud computing (CC) is becoming a necessary technological advancement. This method is a breakthrough in collaborative services and data storage. Nevertheless, the switch to CC has increased security risks, and the networks and daily interactions we engage in depend on network security. An efficient intrusion detection system is essential as attackers create new attack types and network sizes continue to rise. IDS is dependent primarily on determining whether network packets are malicious or benign. Deep learning algorithms have proved to be effective in detecting intrusions compared to other machine learning methods. In this study, we created deep learning methods to recognize attacks using recurrent neural network (RNN) architecture, namely the GRU (Gated Recurrent Unit) architecture. We use these models to handle binary and multiclass classification on the updated cybersecurity CSE-CIC-IDS2018 dataset. The recommended approach offers superior intrusion detection performance regarding Recall, accuracy, and precision. The recommended procedure yielded accuracy and precision values of 99.92 and 99.685, respectively.

Breast cancer is one of the most common diseases and the second most dangerous and fatal disease after skin cancer. The malignant masses begin to grow in the breast cells and develop in the breast tissue of women all over the world. Due to the necessity of early diagnosis to prevent the development of cancer and thus improve the chances of survival, many computer-aided methods are available to automate tissue classification to reduce the workload of the pathologist and improve accuracy. In this paper, a proposed method Gabor-EfficientNetV2 to extract deep features from a breast cancer dataset called “Breast cancer Histopathological dataset” (BreaKHis) to classify it as “benign” or “malignant” based on Gabor filter bank and EfficientNetV2 architecture. The proposed Gabor-EfficientNetV2 model offers significant advances in breast cancer detection and diagnosis by combining the texture analysis capabilities of Gabor filters with the efficiency and scalability of the EfficientNetV2 architecture. This hybrid approach enhances feature representation, resulting in improved classification accuracy and robustness compared to traditional models. Highest classification accuracy obtained with 400X magnification factor where the train accuracy of 97% and train loss of 0.13% while the test accuracy of 96.3% and the test loss of 0.135%, precision of 97.30% , recall of 90.30% , and F1-score obtained is 98.52%.

Because of the fast growing in network system, many categories of intrusion has been discovered that differs from current one and convention firewall and definite rules set and strategies are unable of recognizing this intrusion in real-time. Hence, this demand is requirements of real-times intrusions detection systems (RTs-IDS). The vital aim of this paper is to build an RT-IDSs proficient of classifying intrusion by analyzing the outbound and incoming networks information in real-times. The suggested method contains of deep neural networks (DNNs) trained by use 28 types of the NSL-KDDs datasets. Furthermore, it comprises the machine learning (MLs) pipelines with successive modules for category of data encode and features scaling, that is use before transmit the real-times information to the train DNNs models to create prediction. Composed of the train DNNs models, the MLs pipelines are introduced in the servers that can be access through representation state transfer applications program interface (RESTs API). The DNNs has displayed outstand test performance result realizing around 70% to 96% for f1-score, accuracy, precisions, and recalls. These works comprise a complete practical clarification regarding the implementations and functional of the whole systems. The suggested system usability and efficiency have been increased by its comfort of implementations and remotely accessing. In addition, the proposed model is extremely beneficial for rapidly detects the intrusion by analyze incoming and outbound networks traffics.

Wireless sensor networking is a promising technology with a wide range of applications, including health care and defense. The security of Wireless Sensor Networks (WSNs) is a major concern, particularly for applications where confidentiality is of the utmost importance, because, despite their attractive features (e.g., low installation cost, unattended network operation), these networks do not have a physical line of defense. So, to run WSNs securely, it's important to identify breaches before attackers damage the network or the data sink or base station, which are the nodes that collect and store information. Enhancing the intrusion detection system was suggested in this study using a recurrent Convolutional Neural Network (RCNN) optimised using Particle Swarm Optimisation (PSO) and Coral Reef Optimisation (CRO). Successful intrusion detection has been achieved using the suggested method. When combined, PSO and CRO greatly enhance the efficacy and precision of threat identification. Everyone provides a novel and very effective method to resolve the vulnerabilities that endanger critical infrastructure and safeguard WSNs against them.

The development of fourth generation (4G) and fifth generation (5G) technologies has caused great havoc in favor of telecommunications, leading to new areas of application. These technologies, which provide very high-speed data transfer, expand their range of services from the user's perspective. It is very important to thoroughly investigate the internal mechanisms used by current and modern cellular networks, involving the use of signaling protocols and routing components. Several issues related to the functioning of cellular networks are the subject of many scientific publications. For that reason, this paper does not consider this network, with a focus on the routing process as redundant. The subject is limited to existing cellular systems, which include both fourth and fifth generation technologies, namely 4G and 5G.
A B S T R A C T
The four key features of this review are the terminological changes that have taken place with the evolution of telecommunications technologies, as well as routing mechanisms in cellular networks, next-generation cellular network challenges like vehicle networks, future cellular networks, network slicing, and multi-access edge computing. We characterized our work by describing the inconsistencies observed in the literature on cellular systems. Similarly, this paper identified specific limitations of next-generation cellular systems. In a similar vein, this paper has characterized the 4G and 5G technologies in next-generation cellular networks. Recent attempts at changes in routing have addressed problems related to signaling protocol operations and concerns of future cellular networks. Typically, three criteria are evaluated: low behavioral cost, lack of network congestion, and appropriate route rate. These aspects are mainly considered in the first phase of route research, and the search for the optimization scheme of relevant cellular systems has deepened.

The growth of social media for self-awareness and self-expression has raised more attention to the Myers-Briggs Type Indicator (MBTI) for analyzing people's personalities. Nevertheless, additional research is required to determine various word embedding and data handling methods to enhance the accuracy of MBTI personality-type predictions. Therefore, a new technique known as Random Multimodal Convolutional Forward Taylor Network (RMConv FT-Net) is devised to predict personality. Initially, text data from the database is considered input, and then tokenization is performed to split the text into tokens, which is done by Bidirectional Encoder Representations from Transformers (BERT). After this, various features, including linguistic features and several other features, like Term Frequency-Inverse Document Frequency (TF-IDF), Word2Vec, and so on, are mined from the text. Following this, the data imbalance problem is addressed by employing oversampling. Finally, personality prediction is performed by utilizing RMConv FT-Net, designed with the incorporation of the Convolutional Neural Network (CNN), Taylor series, and Random Multimodal Deep Learning (RMDL). The experimental outcomes of RMConv FT-Net based on Mean Square Error (MSE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE) show that it obtained the values of 0.017, 0.047, and 0.129.

As Internet of Things (IoT) technologies and applications are widely used and spread across various fields such as industry, agriculture, transportation, health, etc., it has become necessary to continuously research and develop information security as technologies and applications evolve. Organizations need to focus their efforts on system security. Any security vulnerability can lead to system failure or cyber-attacks, which can have a wide-ranging impact. IoT security is a protection strategy and defense mechanism that safeguards against the possibility of cyber-attacks that specifically target physically connected IoT devices. IoT security teams are currently dealing with increasing difficulties, such as inventories, operations, diversity, ownership, data volume, and threats. This review examines the research related to security and IoT with a focus on the current situation, applications, issues, and future potential. IoT network security has received increasing attention from multidisciplinary and geographically dispersed researchers in recent years. Data integrity, confidentiality, authentication, and authorization must be ensured due to the large amount of data flowing through network devices. However, the field of IoT security still has a lot of room for growth.

Recent advancements in generative artificial intelligence powered by deep learning have significantly improved image generation and manipulation, resulting in highly realistic images that pose substantial risks to multimedia security. The increasing similarity between authentic and deepfake images, especially those generated by Diffusion Models, highlights the pressing need for effective detection mechanisms. This study provides a comprehensive evaluation of counterfeit facial image detection, focusing on the generalizability and robustness of various detection methods. Using a dataset comprising real images from CelebA-HQ and synthetic images generated by five state-of-the-art models (StyleGAN2, VQGAN, DDPM, PNDM, and LDM), we benchmark four leading detection algorithms: Wang2020, Grag2021, Mandelli2022, and Ojha2023.
The performance of these detectors was evaluated across different generative models and under various image perturbations, such as resizing, noise, blur, and compression. Additionally, we analyzed frequency-domain artifacts, revealing that GAN-generated images exhibit distinct frequency patterns, whereas DM-generated images closely resemble authentic ones. A novel hybrid approach combining spatial and frequency-domain features was proposed, yielding superior performance in detecting AI-generated human faces. Among the methods tested, Mandelli2022 achieved an AUC of 98.38%, while our ResNet-50+FFT model outperformed it slightly with an AUC of 98.42%. These results highlight the effectiveness of hybrid approaches in improving detection accuracy. However, detectors still face challenges in generalizing to diverse datasets, emphasizing the need for more adaptable and robust detection strategies.

As cyber threats continue to escalate with the rapid growth of internet usage, robust intrusion detection systems (IDSs) are essential for safeguarding network infrastructures. This study proposes an enhanced intrusion detection approach using the NSL-KDD dataset, where particle swarm optimization (PSO) is employed for feature selection to optimize machine learning classifier performance. PSO effectively reduces data dimensionality by identifying the most relevant features, improving computational efficiency and detection accuracy. Four machine learning classifiers, such as support vector machine (SVM), decision tree (DT), extra trees (ET), and random forest (RF), are evaluated with and without PSO to assess its impact. Experimental results demonstrate that PSO-based feature selection significantly improves performance, with RF achieving the highest accuracy of 98.33%. Comparative analysis with recent studies highlights the competitive advantage of the proposed method. The study concludes by identifying limitations and proposing future work, including exploring alternative feature selection techniques such as Genetic Algorithm (GA), Bat Algorithm (BA), and Cuckoo Search (CS) to further enhance IDS effectiveness.

This research explores the seamless integration of the Internet of Things (IoT) within the realm of Unmanned Aerial Vehicles (UAVs), resulting in the paradigm of the Internet of Drones (IoD) to create a unified framework that connects both aerial and ground-based systems, enabling an interconnected system. The primary objective is to enhance the operational efficiency, scalability, and innovation potential of drone systems while addressing critical challenges related to connectivity, data management, and security in this emerging paradigm. Securing IoD is essential to protect against a wide range of cyber threats. Moreover, secure communication channels will strengthen the resilience of IoD systems, enabling them to operate safely in dynamic and potentially hostile environments. As IoD systems will rely on continuous communication between drones and IoT devices, securing these exchanges without compromising performance is a primary concern. The research aims to set a foundation for the practical implementation of IoD systems, ensuring secure and efficient operations in diverse applications. The feasibility of implementing a virtual drone framework should be evaluated using open-source simulators or network simulators before deploying the IoD in real-world applications. The Mission Planner simulator, as an open-source tool, offers a wide range of capabilities, including motion capture, collision detection, ease of programming, and support for multiple sensor types, making it ideal for initial tests. In addition, designing a secure IoD communication framework is essential to ensure safe data transfer between IoD endpoints with minimal impact on system performance. To achieve this, the study reviews several secure IoD communication frameworks that incorporate advanced cryptographic techniques. These frameworks are crucial for safeguarding the integrity, confidentiality, and authenticity of data exchanged within the IoD network.

The rapid advancements in artificial intelligence (AI) and data analytics have created significant opportunities in fields such as healthcare and intelligent transportation. As the volume of complex data continues to grow, there is an increasing demand for analytical models capable of extracting meaningful patterns and generating accurate predictions. This study focuses on enhancing Parkinson’s disease (PD) detection by using the Harris Hawk Optimization (HHO) for feature selection to improve classifier performance on the UCI Parkinson's disease dataset. We evaluated four classifiers: Decision Tree (DT), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest (RF), under two scenarios: without feature selection and with HHO-based feature selection. The results reveal substantial performance improvements with HHO, with RF achieving the highest accuracy of 98.33%. Comparisons with recent studies highlight the effectiveness of our approach, establishing it as a new benchmark in PD detection accuracy. This research underscores the essential role of optimized feature selection in enhancing classifier accuracy and reliability, especially for early diagnosis through voice-based data.

This review focuses on the versatility of hybrid models in diabetes prediction, for which early and accurate diagnosis is crucial for patients. Hybrid models have an advantage over traditional approaches since they utilize a combination of machine learning and deep learning to overcome several restrictions inherent in conventional techniques in terms of feature extraction, accuracy, and robustness. Among the structures discussed in this paper, Combine Convolutional Neural Networks and Long Short Term Memory (CNN-LSTM), Support Vector Machine (SVM) with clustering or Decision Tree, and ensemble methods all show high capabilities of capturing the patterns in the diabetes datasets. Analyses state that current typical implementations of hybrid models, intense machine learning, and machine learning achieve the finest steadiness and predictability. However, the following challenges are still experienced: high computational demand, data demands, and interpretability. The subsequent studies should enhance the clinical relevance of these models, including efforts to interpret these models, combine electronic health records, and improve models’ ability to work in real-time before contributing to more effective healthcare solutions for diabetes.

Skin cancer, a dangerous form of cancer, originates from DNA damage that causes unchecked cell growth, resulting in a rapid rise in its occurrence. Earlier research has explored the application of computerised analysis for detecting malignancy in images of skin lesions. Nonetheless, challenges remain, including complications with light reflections, variations in colour, and the diverse shapes and sizes of lesions. This research explores how deep learning techniques can improve the detection of skin cancer. We developed and meticulously evaluated two models: a custom-built Convolutional Neural Network (CNN) and a tailored InceptionV3 model that was pre-trained and adapted for our needs. Our primary goal with the HAM10000 dataset was to execute binary classification, distinguishing between malignant melanoma and benign nevus. The custom Convolutional Neural Network (CNN), noted for its efficient design, achieved a precision rate of 91.80%, while the more complex adapted InceptionV3 model secured an impressive accuracy of 95.72%. The findings showcased here demonstrate the effectiveness of both tailored and pre-trained deep learning models in identifying skin cancer. This showcases their capability to enhance diagnostic accuracy and efficiency, even when encountering various resource and computing constraints. This study highlights how deep learning can significantly improve the accuracy of skin cancer diagnoses, ultimately leading to better outcomes for patients.