بحوث سكوبس — ميسر منذر عدنان

Integrating Wireless Sensor Networks (WSNs) with the Internet of Things (IoT) has transformative potential for data acquisition, processing, and decision-making across dynamic connected environments. Ensuring the security and integrity of these systems is paramount, especially in the face of increasingly sophisticated cyber threats. This study introduces a novel security framework that combines Quantum Key Distribution (QKD) with an adaptive Deep Reinforcement Learning (DRL)-based Intrusion Detection System (IDS), specifically designed to address the unique challenges of the WSN-IoT ecosystem. The key innovation lies in integrating QKD not only for encryption but also as a dynamic quantum-secure layer that continuously adapts to security requirements based on real-time threats and communication patterns. Unlike previous approaches that focus primarily on routing and resource allocation, the proposed framework employs DRL with Proximal Policy Optimization (PPO) to refine intrusion detection by adapting its policies based on evolving attack signatures and threat types. This dual-layer QKD-DRL approach enhances intrusion detection accuracy and establishes a self-optimizing, quantum-secure communication protocol. Tested using the CICIDS2017 dataset, the proposed model achieved a 99.75% detection rate, outperforming traditional Random Forest (97.12%) and Deep Neural Network (96.88%) models. This improvement underscores the efficacy of combining quantum cryptographic techniques with DRL-based adaptive learning, providing a robust, real-time defense mechanism for IoT-driven environments in applications such as smart cities, healthcare, and industrial IoT systems. Thus, the proposed QKD-DRL framework sets a new standard for scalable, secure communication and threat mitigation in the IoT ecosystem. © 2025, Dr D. Pylarinos. All rights reserved.

Attacks on business email compromise (BEC) are becoming more complex and are causing primary data and financial losses. These attacks aim to affect users by sending them emails with incorrect information. Conventional rule-based detection systems fail to spot these attacks.BEC emails are identified in this paper using a machine learning-based approach, a Random Forest Classifier, and email header embeddings. The email headers are preprocessed and converted into vector embeddings to find hidden patterns unique to fraudulent emails. The model is trained on a labeled dataset, including authentic and BEC emails. The proposed method exhibited minimal false-positive outcomes and achieved a detection accuracy of 96.3%. Similar to other baseline machine learning models, it performs much better than rule-based models. It generalizes effectively to various assault variations overlooked previously, supporting the model's longevity. Utilizing embeddings in email header metadata greatly improves the identification of BEC attacks. The Random Forest-based approach provides a real-time, interpretable, scalable solution. This approach proves suitable for strengthening the email security systems of corporate enterprises. © 2025 IEEE.

Fraudulent transactions in business-to-business (B2B) networks create significant dangers that compromise confidence in the financial system and operational stability, typically using rule-based or static machine learning algorithms to capture the temporal complexity of fraudulent activity patterns. The study aims to develop a framework for fraud detection utilizing Extreme Gradient Boosting (XGBoost) with temporal pattern extraction to enhance the prediction accuracy. The approach's fundamental component is preprocessing the transaction data to identify time-dependent characteristics, including periodicity, transaction bursts, and frequency anomalies. It ensures total correctness. These features can be added to the XGBoost model to detect evolving advanced fraud signs. An investigation on a real-world B2B dataset revealed that the model had an accuracy of 92.4% and a recall of 89.7%, well beyond the values reached by baseline models. The experiment was conducted to assess the model's accuracy. Incorporating temporal features greatly enhanced early fraud detection skills. The final result is that the offered approach offers an interpretable and scalable solution for spotting real-time fraud in B2B environments. This approach reduces financial losses and strengthens transaction security. © 2025 IEEE.

Autonomous automobiles need a real-time, consistent pathfinding system to appropriately and efficiently negotiate changing surroundings. Sometimes, traditional approaches fall short in multiple respects, including scalability, flexibility, or handling of difficult terrain. This paper presents an improved Ant Colony Optimization (ACO) method to help nav-assisted autonomous automobiles overcome obstacles. The enhanced ACO adds environmental feedback systems, dynamic heuristics, and adaptive pheromone updates to improve route selection under real-time restrictions. The key goals in urban and off-road settings are to lower calculation time, accident risk, and route distance, thus enhancing pathfinding efficiency. This work has been simulated using reasonable traffic models, and the method is compared to industry standards in navigation systems like A∗ and Dijkstra's to evaluate it. Comparatively to traditional methodologies, the results of trials show an increase in convergence time, an improvement in route optimality, and an increase in adaptability to barrier conditions. The suggested ACO variation provides next-generation intelligent transportation systems with scalability and outstanding performance and boosts the decision-making capability of autonomous systems. The following projects will concentrate on hardware implementation on actual autonomous systems to evaluate their viability and real-time responsiveness under different circumstances. © 2025 IEEE.

The evolution of mHealth apps has offered a crucial substitute for round-the-clock patient monitoring, especially in managing chronic illnesses. However, centralizing sensitive patient data creates significant security and privacy issues. To protect patient information in real-time monitoring systems, this study proposes a secure architecture that combines Federated Learning (FL) with Differential Privacy (DP). Allowing distributed model training across edge devices ensures that local devices retain raw patient data. The FL approach reduces the probability of data leakage from shared parameters by incorporating randomization into model updates that utilize differential privacy. The suggested method was evaluated on a real-world mHealth dataset for health condition prediction, considering both accuracy and privacy loss factors. The results indicate that the model has good privacy guarantees (ϵ < 1.5) and outstanding prediction performance (95.4% accuracy). The suggested method offers acceptable trade-offs between data value and user anonymity, hence offering an attractive option for safe patient monitoring in settings where privacy is vital but resources are limited. Future studies on adaptive privacy budgets and energy-efficient technologies will guarantee their long-term use in wearable healthcare environments. © 2025 IEEE.

Privacy, data security, and detailed access control issues significantly hinder secure electronic health record (EHR) interchange in decentralized healthcare settings. Traditional access policies based on identities or roles might fall short in dynamic, multi-user systems where the diversity of traits is essential. This study aims to provide a more secure access control system by combining attribute-based encryption (ABE) with bilinear pairings, enabling the safe storage of electronic health record data under flexible and scalable access controls. Using Ciphertext-Policy Attribute-Based Encryption (CP-ABE), the approach provides access to encrypted data based on a predefined set of characteristics for healthcare professionals, patients, and insurers. Using bilinear pairings, encryption is theoretically strong and cryptographically resistant. After simulated EHR datasets were used to construct and test the system, its performance was evaluated in terms of processing efficiency, access time, and resistance to unauthorized decryption. Significant findings indicate that the proposed method offers scalable policy enforcement with robust data secrecy and reduces processing costs compared to traditional ABE schemes. Ultimately, ABE with bilinear pairings provides a practical means for the safe and policy-driven transfer of electronic health records (EHRs), hence enabling patient information protection and promoting interoperability across hospital systems. © 2025 IEEE.

The demand for quick development cycles and the growing complexity of software systems support the need for intelligent automation in coding processes. Conventional software development approaches are time-consuming, error-prone, and usually unable to satisfy high standards. This work presents an automated code-generating method based on GPT-based language models to increase software development efficiency. The approach combines prompt engineering techniques and reinforcement learning strategies with fine-tuning a pre-trained GPT model on various high-quality programming datasets to optimize the relevance and accuracy of output. A validation module helps validate the created code in terms of both syntactic and semantic. Based on experimental data, the method reduces development time by as much as 40% compared to hand coding and achieves an average code correctness rate of 88% over several programming languages. With little to no human input, the results show the model can produce code that fits the particular context and is functional. All things considered, GPT-based models, when customized to specific software domains and reinforced with validation systems, can transform future software development methods. They offer a dependable, customized, reasonably priced automated code-generating solution. © 2025 IEEE.

One of the most common uses of computer vision, automatic number plate recognition (ANPR) is also a pretty well-explored subject with numerous effective solutions. Due to regional differences in license plate design, however, these solutions are often optimized for a specific setting. Number plate recognition algorithms are often dependent on these aspects, making a universal solution unlikely due to the fact that the image analysis methods used to develop these algorithms cannot guarantee a perfect success rate. In this research, we offer an algorithm tailor-made for use with brand-new license plates in Iraq. The method employs edge detection, Feature Detection, and mathematical morphology to find the plate; it was developed in C++ using the OpenCV library. When characters were found on the plate, they were entered into the Easy OCR engine for analysis. © Copyright Alsudani.

In this article, we looked at how to go about creating a CNC pen or drawing machine of your own. Inkscape, which translates graphics and text into g- code format, was utilized as the controller for this project, with the microcontroller serving as the interface between the computer and the language of the CNC machine. The g-code transmits a series of x, y, and z coordinates to the motors; the servo motor controls the pen's movement in response to the Z coordinates; stepper motor 1 controls the rail's horizontal motion; and stepper motor 2 controls the rail's vertical motion in response to the X coordinate. The laser machine employs industrial applications to expedite manufacturing and perform engraving and cutting, resulting in a superior and expertly finished output. The carbon laser beam emitted by the laser engraving machine may be used for engraving, cutting, and shaping a wide variety of materials and end products. ©Copyright Yousif.