AL-Rafidain Journal of Computer Sciences and Mathematics

An Advanced GAN-Based Framework for Medical Image Enhancement

Saja Younis Hamid Alhamdani — 2025-12-25

Low contrast, noise, and low visual detail are major medical image problems that pose a negative effect on diagnostic accuracy. This paper suggests a more developed model using deep generative networks (GANs) to enhance the quality of medical iris images. The framework has a sequence of preprocessing steps that include contrast enhancement (CLAHE), noise removal (Bilateral Filter), and edge enhancement (Unsharp Masking), and then the stage of enhanced generation with an attention-assisted generator (Adam) with fine-tuned parameters. SSIM, PSNR and LPIPS measures were used to evaluate the performance of the model. The findings revealed that there were significant visual and perceptual structure of images as results showed that, average SSIM was improved by 0.9383 to 0.9783, LPIPS was reduced by 0.0137 to 0.0078 and PSNR had increased by 28.62 to 32.23 than the default parameters. These results confirm the usefulness of fine-tuning at enhancing perceptual and structural image measures. This model improves the diagnosis abilities in the medical field and minimizes the use of costly refined imaging methods, hence it can be applied in large scale clinical setting.

Quasi Triple Operator of Order

Arwa Nazar Mustafa — 2025-12-25

In this study, two new classes of operators on a complex Hilbert space H were presented and referred to as the Quasi-triple operator and quasi-triple operator. Quasi-triple operator is denoted as qt-operator, and the quasi-triple operator is denoted as qt-operator where is a linear bounded operator on H. The generalization of the above concepts is Quasi-triple operator of order n and quasi-triple operator of order n. Some results of these two types of operators including sum, product, direct sum and tensor product operations have been discussed. We also presented some results and theorems and supported the discussion with some illustrative examples.

Development of a Spectral Conjugate Gradient Method for Solving Optimization Problems

Kamal Jameel Murad — 2025-12-25

Conjugate gradient methods have been favored to use for their efficiency in solving large-scale unconstrained optimization problems, primarily because of their low memory requirements and exclusive to use the first-order derivative information. In this paper, we introduce a spectral conjugate gradient method that enhances the classical approach by merge a spectral property directly into the determination of the search direction. At the core of our method lies a developed formulation of a spectral search direction and a more precisely adjusted conjugate gradient coefficient, both derived as extensions of established conjugacy condition. To ensure numerical stability, we also include a correction term that accounts for the limitations of machine precision. Our theoretical analysis confirms that the developed method generates search directions satisfying the descent condition, which is critical for ensuring convergence. To assess its real-world effectiveness, we subjected the spectral conjugate gradient method to an extensive set of numerical experiments and benchmarked its performance against that of a standard conjugate gradient method. By using range of test problems, our method consistently delivered superior results, particularly in reducing the number of function evaluations and exhibiting improved scalability in higher-dimensional settings. These findings strongly indicate the spectral conjugate gradient method’s potential as a reliable and efficient tool for optimization. Future research may explore further refinements to the method’s theoretical foundations, investigate its performance in constrained or stochastic environments, and apply it to practical optimization challenges such as neural network training, signal recovery, structural design, and control system calibration.

Fault-Tolerant SCADA Systems Using ECC for Enhanced Cybersecurity and Data Integrity

Iehab Abduljabbar Kamil — 2025-12-25

This study focuses on the implementation of Error Correction Code (ECC) in Supervisory Control and Data Acquisition (SCADA) systems for a better performance of error controls and the system as well. Since the SCADA systems are fundamental in supervising industrial processes, the study focuses on the issue of error control to avoid interferences. The study proposes the Integrated Control Flow Checking or ICS-CFC methodology, which increases the reliability of SCADA systems to neutralize errors with considerably minimal overhead costs. In the critical trials performed on various simulated IT infrastructures and real ICS of industrial organizations, the proposed ICS-CFC achieved a fault coverage of 96.32% and fairly reasonable memory and performance overheads as well. The lack of additional hardware needed to implement the methodology makes it inexpensive besides enhancing already existing SCADA systems. Therefore, the analysis finds out that ICS-CFC enhances the error handling capability and reliability of SCADA and can be considered a workable solution for industries with stringent and consistent and occasional error requirements. Thus, for further ECC methods application for more variants of SCADA systems, as well as to improve operational and security features, future work is suggested.

The Zagreb Coindices to the Zero Divisors Graph of Principal Ideal Local Rings

Ahmed S. Ismail — 2025-12-25

A Hybrid Conjugate Gradient Algorithm Using the Golden Section Ratio for Unconstrained Optimization Problems

Sara Sahib Mohammed Zaki — 2025-12-25

Solving Reliability-Redundancy Allocation Problems using Grasshopper Optimization Algorithm

Mafaz Emad Abid Salih — 2025-12-25

Success of companies and satisfaction of customers heavily depend on system reliability performance. System design improvements and higher efficiency exist directly from the proper distribution of reliability overhead resources. The extensive application domain of Reliability Redundancy Allocation Problems (RRAPs) includes fundamental challenges in various real-life situations such as software design along with cost optimization and development. The system reliability optimization problem is recognized as NP-Hard, and its resolution demands planned and effective solution approaches since there doesn’t exist a polynomial-time method for finding optimal solutions. This research utilizes Grasshopper Optimization (GOA) due to its ability to solve complex constrained optimization problems effectively and its high accuracy in obtaining good solutions. Eight system reliability block diagrams were used, varying in their difficulty from simple to complex problems. Results showed that system reliability comprised a significant increase when GOA-based optimization was applied compared to other algorithms. GOA achieved higher performance using the eight system reliability block diagrams, and its results validated both the efficiency and the solution-delivering capabilities of the algorithm for enhancing overall software reliability.

Darboux Integrability Of the Biological System

Zakariya Hashem Ali — 2025-12-25

In the given paper, we investigate the integrability of a mathematical model of a 3D biological system. Our results show that the system admits a polynomial first integrals for some parameters, an invariant algebraic surface with an exponential factor, and Darboux first integral. The proof was realized with the help of weight homogeneous polynomials. A model combining virus therapy and chemotherapy holds promise for improving the efficacy of cancer treatments. Virus therapy can target and destroy cancer cells, while chemotherapy can enhance the immune response and sensitize the tumor to viral therapy.

A Survey on Federated Learning: Fundamentals, Challenges and Client Selection Methods

Alaa Amjad Mala Baker — 2025-12-25

Federated learning (FL) represents a transformative shift in machine learning, moving from conventional centralized approaches to a distributed framework that emphasizes data privacy and security. The FL server transmits an initial model to clients, which they train locally on their private data. After training, the server aggregates model updates from each client to update the global model. Selecting the best clients in FL is critical to improving the convergence speed and accuracy of the final model, which requires careful client selection approaches. The client selection phase of FL faces numerous challenges that impact overall training performance, including statistical heterogeneity and system heterogeneity resulting from the diversity of client data and resources. Communication costs present another challenge, especially in networks with limited client communication resources. Additionally, selecting trustworthy clients represents another challenge, as selecting malicious clients creates a significant risk within the FL training process. Moreover, the fairness challenge entails providing fair opportunities for all clients to participate in training. To address these challenges, we offer solutions that utilize effective techniques, approaches, and client selection methods. This survey presents a taxonomy of modern client selection methods in FL, highlighting the improvements in FL performance and effectiveness achieved through these methods, including greedy selection, reinforcement learning-based selection, multi-armed bandit-based selection, clustering-based selection, and reputation & security-based selection. Subsequently, it offers a general comparison between these methods in terms of their core ideas, advantages, limitations, and use cases. Finally, future and potential trends in client selection and improving performance in FL are identified.

Review of a Unified Real-Time IDS and Mitigation Framework Using Apache Spark

Ammar Ahmed Abdullah — 2025-12-25

Traditional intrusion detection systems are being surpassed by the increasingly sophisticated cyber threats that modern networks face. The increasing scale and complexity of modern network environments, coupled with the evolving sophistication of cyber threats, have rendered traditional Intrusion Detection Systems (IDS) inadequate for real-time and large-scale protection. This paper presents a comprehensive review and design strategy for a unified, real-time IDS and mitigation framework leveraging Apache Spark. This paper proposes a unified real-time IDS framework that utilizes Apache Spark to address the aforementioned disparity. The design combines threat intelligence, distributed machine learning, and streaming data analytics to facilitate automated mitigation and scalable multi-vector threat detection. We have identified critical limitations (e.g., offline detection, limited attack scope, outdated datasets) and have developed a set of objectives to address them through a review of current Spark-based IDS research. The outcome is a definitive roadmap for a next-generation IDS that offers low-latency, adaptive, and transparent defense in high-throughput network environments.