Al-Furat Journal of Innovations in Electronics and Computer Engineering

Artificial intelligence (AI) and deep learning have recently attracted a lot of attention as potential tools for the early detection of ocular diseases. Reason being, these innovations have the potential to substantially enhance therapeutic options and healing rates. With the use of algorithms built using massive medical datasets, a number of common eye disorders may now be detected automatically. These include myopia, diabetic retinopathy, macular degeneration, cataracts, and glaucoma. On a worldwide basis, cataracts and glaucoma are two of the leading causes of permanent visual loss. However, testing the viability of AI for the diagnosis of these two illnesses has been sparse in comparison to other age-related eye problems. An extensive review of deep learning algorithms for glaucoma and cataract detection using slit-lamp and fundus images and other relevant data is presented in this paper. As an added bonus, it explores state-of-the-art algorithms and models, preprocessing methods, retinal image databases, and performance evaluation standards. The main foci of this research are cataract diagnosis and glaucoma detection and classification. via the dissemination of new scientific knowledge to researchers and ophthalmologists via this comprehensive review, this study aspires to contribute to the progress of ophthalmology and the field overall.

Noise is undesired information in an image that appears during image capturing, transmission, coding, and processing steps, negatively impacting visual quality and obscuring important details. Image denoising is therefore a central problem in both image processing and computer vision, aiming to suppress such distortions while preserving essential structural information. Our paper aims to shed light on recent research in image denoising techniques, evaluate the performance of prominent methods, highlight research gaps, and identify areas for further study.

Encryption algorithms such as the block cipher techniques were built in particular to guarantee security on limited systems, whose design clarity is vital. The multiplicity of the planned use cases, however, needs flexibility in execution. In the building of cryptosystems, simplicity, security, and flexibility are continuous but incompatible aims. The requirement for a lightweight block cipher algorithm that supports a broad range of systems, architectures, and block/key sizes efficiently has been developing recently. In 2013, the National Security Agency (NSA) recommended the block ciphers SIMON and SPECK. Different block sizes, including 16, 32, and 64 bits, and key sizes, such as 64, 96, and 128 bits, are supported by SIMON and SPECK. In this research, we evaluate the performance of lightweight SPECK cryptographic block ciphers utilizing multiple criteria such as execution time, throughput, and energy usage. We utilize two distinct Intel CPUs to assess the Speck Cryptography technique. Three keys of sizes 128, 192, and 256 are employed in this article to compare the performance criteria. The obtained results over two different Intel processors show a linear and non-linear relationship to the key size. In other words, the increase in key length increases the execution time and energy consumption, while throughput results decrease. Moreover, the NIST statistical randomness test is implemented for all speck versions to compare them in term of the security level.

Drones are currently being used more and more to deliver postal goods across short, medium, and long distances. Unmanned aerial vehicles (UAVs), especially those with vertical take-off and landing (VTOL) capabilities, have the potential to be used in transportation services due to almost unrestricted access to areas without significant airspace errors. So drones are very important. Part of the process of implementing transportation services. These days, the worldwide Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS) help UAVs throughout the flight phase. However, GNSS+INS's imprecision renders it irrelevant for landing with take off, entailing the assistance of human drone operators along such stages. Unfortunately, the accuracy of the navigation technologies currently in use is insufficient to locate drones. Because of this, full automation of the landing technique is Not feasible. Since the study describes the design of the autonomous system, this paper offers a solution to this issue. Automatic landing approaches are made possible by Doppler-effect-based navigation algorithms as a sophisticated accuracy evaluation. Simulation studies serve as the foundation for the vertical take off and landing (VTOL) solution. The proposed frequency hopping modulation with the Doppler effect adaptive control model has been designed and implemented to simulate the successful landing of a typical drone structure. Experimental results have shown excellent metrics for frequency hopping signal of (20-50) MHz carrier frequency and 150 hops/sec specified to UAV speed at 80 m/s having Doppler effect with N=1024 and 2024 samples. The obtained results show that the received control signals for Doppler effects match the results of the adaptive compensator, which indicates an enhancement in the take-off and landing operations of the drone model. The observed improvement was up to 98%, with an error rate of less than 0.2%, with reception power ranging from 40 to 100 dB at the operating frequencies of the path model.

This research examines the execution of feed-forward artificial neural networks on field-programmable gate arrays (FPGAs), addressing the challenge of limited hardware resources that restrict network size and hinder the direct execution of nonlinear activation functions. The proposed approach reduces resource usage through data normalization and restructuring, while approximating the tan-sigmoid function via a logarithmic method. Implemented on the Xilinx Spartan-3A (3sd3400afg676-4) platform, the design achieves a 3% optimization in slices and a 2% optimization in lookup tables (LUTs), demonstrating improved efficiency, enabling larger and more cost-effective neural networks on a single FPGA chip.

Face anti-spoofing (FAS) has lately received more attention because of its critical function in protecting face recognition systems against presentation assaults (PAs). With the advent of more realistic face recognition systems of various sorts, early-stage face anti-spoofing approaches based on hand-crafted characteristics have become untrustworthy due to their limited representativeness. With the advent of large-scale academic datasets over the last decade, deep learning-based face anti-spoofing algorithms have outperformed and dominated the area. However, most existing evaluations in this field rely on hand-crafted features, which are now outdated and hinder progress in the face anti-spoofing research community.
To inspire future research, we give the first complete analysis of recent advances in deep learning-based face anti-spoofing systems. This article addresses numerous innovative and significant components: 1) In addition to binary label supervision (e.g., "0" for truth versus "1" for face recognition systems), we also examine cutting-edge approaches using pixel-level supervision (e.g., pseudo depth map). 2) In addition to traditional evaluation within the dataset, we collect and analyze cutting-edge methods specifically designed for domain generalization and open-set FAS analysis; and 3) In addition to commercial RGB cameras, we summarize deep learning applications under multimodal (e.g., depth and infrared) or specialized (e.g., light field and flash) sensor categories. We end our survey by stressing existing outstanding challenges and future opportunities.

With the continuous development of the cybersecurity field, detecting unknown or zero-day attacks remains a significant challenge due to their unpredictable behavior. This paper proposes a novel biologically inspired approach to detecting unknown SQL Injection (SQLi) attacks using the Slime Mould Algorithm (SMA). This method leverages the adaptive and heuristic capabilities of the SMA to detect unknown attacks. A binary classification model was developed and trained on a benchmark dataset containing both natural queries and diverse SQLi attack vectors, including Out-of-Band, Boolean-based, Time-based, and Union-based injections. To ensure robustness and generalization, K-Fold cross-validation was employed. The SMA-based model demonstrated superior detection capabilities, particularly in identifying zero-day attacks that deviate from known patterns. The experimental results reveal promising detection rates across all attack types: 89.33% for Out-of-Band-based SQLi, 97.89% for Boolean-based SQLi, 90.27% for Time-based SQLi, and 96.69% for Union-based SQLi, and 91.51% for Error-based SQLi. These results underline the effectiveness of SMA in generalizing beyond seen data, a critical advantage in dynamic threat environments. Compared to traditional machine learning models, the SMA-based classifier achieved higher accuracy and F1 scores, confirming its potential as a powerful tool for web application security.

Visible light communication (VLC) has appeared as an attractive wireless technology for indoor data transmission, but its performance is restricted by such a fluctuating environment as the indoor channel caused by obstacles, reflections, and ambient sunlight interference. The conventional direct current-biased optical generalized frequency division multiplexing (DCO-GFDM) requires a Hermitian symmetry to generate real values, which are restricted by the light source. Hence, the size of the IFFT/FFT and the pulse shaping filter must be doubled, which leads to higher complexity and power consumption. This paper introduces a machine learning (ML)- based adaptive strategy for non-Hermitian DCO-GFDM (NH-DCO-GFDM), which was proposed in our prior study. This strategy is developed to change the modulation parameters according to the channel conditions dynamically. We employ random forest regression models trained on simulation-generated datasets to predict critical performance metrics, including bit error rate (BER) and peak-to-average power ratio (PAPR). The proposed algorithm works to reduce the BER under a specific threshold. Then, it optimizes the modulation parameters via a cost function that maximizes throughput while controlling complexity and PAPR. The results indicate that this adaptive system significantly improves reliability and achieves BER below the threshold at a lower SNR (9 dB), providing higher throughput. Thus, the proposed method offers an effective solution for robust, high-throughput VLC that can trade with such dynamic indoor environments.
Keywords: Machine learning, Random Forest regression, GFDM, VLC,

The negative aspects of the academic social networks, such as ResearchGate (so
shallow measures as co-authorship and citation numbers), is where the present paper proposes a
hybrid data analysis model in which more valuable data sets of scholarly collaboration at University
of Kufa are pursued. Research A network was created among scholars who were interested in
research at the University of Kufa. To identify semantic clusters, the K-Means clustering algorithm
was used on the TF-IDF vectors using K-Means and dimensionality reduction using SVD, and
structural communities using the Louvain algorithm. The most suitable value of k was assessed with
the help of the Elbow method. Findings indicated that the K-Means clustering performed better and
the Louvain worse on maximum Silhouette score and Calinski-Harabasz score and integrated
semantic- cohesion in research interest networks. These will be the structural and semantic analysis
tools that will give additional measurements and the Louvain and K-Means functions that will
determine the topological form and theme relations respectively when applied to the semantic
material. The article warrants the existence of hybrid analytic models, feasibility of designing
comparative academic suggestion systems, which have structural location as well as semantic
proximity.

Wireless Sensor Networks (WSNs) are widely used in various applications, yet they remain limited by the energy constraints of sensor nodes. This review paper explores and compares recent advancements in energy-efficient routing strategies across four major categories: Q-Learning-based routing, traditional optimization algorithms, clustering-based protocols, and sleep scheduling techniques. Each approach is analyzed in terms of its methodology, simulation environment, performance metrics, and limitations. Q-Learning techniques provide adaptive and intelligent routing decisions but often lack real-world deployment. Traditional algorithms such as Ant Colony Optimization (ACO) and Whale Optimization Algorithm (WOA) offer reliable clustering but adapt poorly to dynamic environments. Clustering-based protocols, especially those integrating fuzzy logic or quantum methods, show strong results in simulations but assume static and homogeneous nodes. Sleep scheduling and duty cycling protocols significantly reduce idle energy waste, yet are rarely integrated with routing layers. The review identifies that most current protocols are evaluated only through simulations and face challenges such as congestion near sinks, lack of cross-layer integration, and scalability under heterogeneous conditions. Future research should focus on building lightweight, real-time learning frameworks that jointly optimize routing, clustering, and sleep scheduling in practical deployments.

The goal is to compare deep learning and classical methods for automated ECG beat classification in a rigorous inter-patient evaluation. Band-pass filtering, baseline correction, Kalman denoising, and QRS detection using the Pan-Tompkins algorithm were applied to 48 half-hour, two-lead recordings from 47 subjects from the MIT-BIH Arrhythmia Database. Discrete wavelet transforms features fed a tree-based baseline (random forest, RF) following R-peak segmentation were compared to a convolutional neural network (CNN). Per-class precision/recall/F1, macro-F1, and ROC-AUC/PR-AUC with 95% bootstrap CIs were reported. In a four-class setup (atrial fibrillation, normal, other rhythms, and noise), RF obtained macro-F1 ≈ 0.70 and CNN macro-F1 ≈ 0.59. AF↔Other, Noise↔Other, and Noise↔AF were the most commonly confused terms. In inter-patient evaluation, RF serves as a competitive baseline. Comparability across studies is improved by PR-centric evaluation, standardized reporting, and clear preprocessing, especially when there is a class imbalance.

In this paper, a novel design of a multifunctional quasi-reflectionless non-reciprocal diplexer is proposed. The reflection less diplexer is necessary to overcome many problems in various parts of communication systems that result from reflective unwanted signals from passive elements. in addition, the nonreciprocal diplexer has many advantages in RF systems as interference elimination. The proposed diplexer consists of two pseudocombline BPFs that work in two channels (2.2, 2.6) GHz, respectively. A T-shaped structure is used in the proposed QR diplexer design in three parts (an absorptive part for each BPF and a matching part between the two BPFs). Also the Spatial Temporal Modulation (STM) is used for achieving the nonreciprocity property for the diplexer. Furthermore, the proposed model achieved a high level of absorption for undesired reflected signals, where undesirable waves in the 1.5–4 GHz frequency band were efficiently absorbed. The proposed QR diplexer demonstrates a wide range of suppression for unwanted signals as well as significant separation between the RX and TX ports. All simulations are performed by Keysight ADS software and fabricated by a CNC machine. The results of the proposed work show good agreement between simulation and fabrication.

The global healthcare sector stands at a critical juncture, facing unprecedented challenges including aging populations, rising costs, workforce shortages, and the environmental burden of medical waste. Concurrently, it is experiencing a digital revolution, generating vast repositories of data, particularly medical images which constitute nearly 90% of all healthcare data. This paper posits that the strategic integration of Artificial Intelligence (AI) and advanced medical image processing technologies is not merely an incremental improvement but a foundational paradigm shift essential for achieving sustainable, data-driven healthcare. We move beyond technical performance metrics to explore a holistic framework where AI-driven imaging enhances all three pillars of sustainability: economic (through improved efficiency and reduced waste), social (through equitable access and improved outcomes), and environmental (through optimized resource use). This paper provides a detailed analysis of current applications in radiology, pathology, ophthalmology, and cardiology, supported by expanded clinical examples and data tables. We critically examine the "green" implications of computational costs, the ethical imperatives of bias mitigation, and the practical pathways for implementation. Ultimately, we argue that an AI augmented, image centric approach is the most viable pathway to a healthcare system that is not only more accurate and efficient but also profoundly more sustainable and equitable for future generations.

For more than a decade, the need to share the frequency spectrum between radar and wireless communication systems has emerged due to the massive increase in demand for communication services and the availability of underutilized radar bands. At the same time, spectrum sharing between radar and wireless communication systems faces the problem of interference that affects the performance of both systems, which has attracted researchers to find effective solutions to reduce interference in an environment where both systems share the radio spectrum. In this paper, we review the most prominent optimization methods that have been used to share the frequency spectrum and reduce interference between radar and wireless communication systems.

This study highlights the design, simulation, and performance assessment of a standalone photovoltaic system developed for a rural educational facility in Babil, Iraq, under actual site conditions, the work fulfilling a novel weekday only for the electrical requirements of Al-Nasra Primary School, situated adjacent to the Al-Musayyib Grand Irrigation Project. The system was modeled and analyzed using PVsyst software, utilizing a realistic weekday-only load profile that simulates school operations for five hours daily (08:00–13:00), five days a week. The entire daily energy requirement was assessed at 62.0 kWh, encompassing classrooms, administrative offices, and heating and cooling systems. The results of the simulation indicated that the designed system produced a yearly energy output of 21,306 kWh, with 15,927 kWh effectively employed to meet the load and 4,295 kWh recorded as surplus energy. The system showed the performance ratio (PR) of 62.0%, solar fraction of 99.72% and a load loss of 0.6%. The battery bank maintained a state of health exceeding 91.7%, with an estimated lifespan of 12 years. These results verify the technical viability and dependability of a suitably sized standalone photovoltaic system in fulfilling the distinct operating requirements of a rural school. Furthermore, the setting up of a pragmatic weekday-only load profile yields a more precise evaluation of system performance, preventing oversizing and promoting improved energy management strategies for schools in off-grid or locations with an unsteady power grid. The weather data employed for the simulation was obtained from the PVsyst linked Meteonorm 8.0 data for the Al-Musayyib region.

Wireless Local Area Networks (WLANs) are one of the fastest growing technologies
today. In recent years, WLANs have become one of the most adopted technologies. Nevertheless, wireless
communications presents a huge security risk. As WLAN is susceptible to many attacks, encryption
methods are the building block of WLAN security. However, it can be difficult to develop a practical
system that provides high security but also has decent performance. The research proposes a new solution
for a dual hybrid encryption method namely Advanced Encryption Standard with Elliptic Curve
Cryptography (AES–ECC) and Advanced Encryption Standard with Rivest–Shamir–Adleman (AES–RSA).
The paper compares two hybrid approaches to develop a unified framework. So, the key novelty is to
combine the two approaches under the same framework. In the method proposed, symmetric encryption
(AES) secures the sent data while asymmetric encryption (ECC or RSA) encrypts and securely sent an
important secret key. The main contribution of this paper is to offers equal characterization of AES–ECC
and the AES–RSA hybrids in WLAN environments as well as to provide an unbiased comparison of their
results and security, resources usage and processing time. Experimental testing shows how effective and
practical this framework is for high-level security and performance of WLAN digital communications.

This research integrates the Hill Cipher encryption algorithm with DNA-based steganography as a form of dual protection. The plugins and modules mask the cryptex as DNA sequences. The proposed methods presented the full accuracy of the data recovery with low time and full recovery. The encryption and decryption steps are of low time and full recovery. The effective masking and absence of the cryptex within the cover DNA, as nominal 50%, and the absence of the cryptex confirm the absence of cryptex and detection. The proposed model gives an extremely high potential for secure communication, in high confidentiality environments, like in the exchange of medical and governmental data.