Iraqi Journal for Computers and Informatics
Not a member yet
273 research outputs found
Sort by
A Dataset-Driven Comparison of Traditional and Advanced Machine Learning Techniques for Phishing Detection in Low-Variance Environments
No full textPhishing attacks continue to grow rapidly, often using obfuscated and deceptive URL patterns to mimic legitimate websites and evade detection. While traditional Machine Learning (ML) models perform well on benchmark datasets, they often struggle in real-world scenarios where phishing URLs are carefully crafted to resemble authentic domains. This study presents a dataset-driven comparison between traditional ML models—Logistic Regression, K-Nearest Neighbors, Support Vector Machine, and Random Forest—and advanced approaches such as Extreme Gradient Boosting (XGBoost), a tuned XGBoost variant, and a soft voting ensemble. Two datasets were used: (i) a high-variance global dataset from Mendeley Data, and (ii) a custom-built local dataset with region-specific phishing URLs designed with minimal alterations (e.g., character substitutions and deceptive subdomains) to simulate low-variance attacks. Preprocessing included feature engineering, variance analysis, and balancing with the Synthetic Minority Oversampling Technique (SMOTE). Experimental results show that Random Forest outperforms other traditional models but still struggles with low-variance phishing URLs. In contrast, advanced models—particularly tuned XGBoost—achieved significantly higher recall (0.99) and strong precision (0.81), while the voting ensemble further improved robustness by combining multiple classifiers. These findings emphasize the importance of realistic datasets and demonstrate that advanced ML strategies are more effective for detecting phishing attempts based on subtle obfuscation. This work contributes by (i) validating advanced ML models under realistic low-variance conditions, and (ii) highlighting precision and recall as more appropriate evaluation metrics than accuracy in cybersecurity
CNN-Based Transfer Learning Approach for Cross-Platform IoT Malware Detection
No full textThe increasing prevalence of Internet of Things (IoT) devices has raised significant concerns regarding their security. Malware attacks on these devices can lead to severe consequences, including data breaches, privacy violations, and system failures. This paper proposes a novel approach for detecting malware in cross-architecture IoT devices using deep learning with Convolutional Neural Networks (CNNs). The proposed methodology involves converting binary files into grayscale images and utilizing a CNN model for feature extraction and classification. The model\u27s performance was evaluated on a dataset comprising malware samples from various IoT devices, achieving an accuracy of 97%. The results demonstrate the effectiveness of the proposed approach, outperforming existing methodologies in detecting malware in cross-architecture IoT environments. The model\u27s robustness and adaptability across various malware samples highlight its potential as a valuable tool for enhancing IoT security. Future work will focus on expanding the dataset to incorporate more diverse and complex malware samples and exploring additional deep learning architectures
Cross-sectional time series models for identifying the most important factors of e-government growth in Arab countries
Full text linkThe goal of this study was to determine the key elements driving the expansion of e-government in Arab nations. One of the econometric models—the cross-sectional time series model (panel)—was employed to accomplish the intended outcome. The following techniques were used to estimate the models: fixed effects model, random effects model, and linear panel data regression analysis. The results revealed that e-government is negatively affected by political stability and the absence of terrorism and violence (PV), positively affected by government effectiveness (GE), positively affected by regulatory quality (RQ), and positively affected by the rule of law (RL)
Automated Object Detection and Count Estimation Based on Machine Learning Models
No full textObject detection and counting is a crucial problem in various areas like autonomous machines, health sector and industrial automation. This paper presents a comparative study between two Machine Learning (ML) methods, depending on Single Shot Multi Box Detector (SSD) with You Only Look Once (YOLOv3) and MobileNetv3 to focus on the object detection and counting. The well kown dataset (COCO) used to evaluate the two models where 44 images chosen randomly for testing. In addition SSD and MobileNet v3 compared in a real time mode. The performance metrics used in this paper were confidence scores, average over image and processing time. The results indicate that there was an tradeoff between accuracy and speed. Both models showed a fast inference time, which made it suitable for real-time applications, although (YOLOv3) was more confident in some cases because of its complex architecture. The performance difference is caused by the design of the model: MobileNet\u27s simple structure aims at lightweight, while (YOLOv3)\u27s complex network enhance robustness in detection. This work highlights The necessity for application-aware model selection balancing between the fast performance on edge devices (such as drones) versus accuracy for accuracy-precision applications (like medical imaging). The findings provide practical insights for deploying ML-driven object detectors across various domain
LLM Hallucination: The Curse That Cannot Be Broken
Full text link
Artificial intelligence chatbots (e.g., ChatGPT, Claude, and Llama, etc.), also known as large language models (LLMs), are continually evolving to be an essential part of the digital tools we use, but are plagued with the phenomenon of hallucination. This paper gives an overview of this phenomenon, discussing its different types, the multi-faceted reasons that lead to it, its impact, and the statement regarding the inherent nature of current LLMs that make hallucinations inevitable. After examining several techniques, each chosen for their different implementation, to detect and mitigate hallucinations, including enhanced training, tagged-context prompts, contrastive learning, and semantic entropy analysis, the work concludes that none are efficient to mitigate hallucinations when they occur. The phenomenon is here to stay, hence calling for robust user awareness and verification mechanisms, stepping short of absolute dependence on these models in healthcare, journalism, legal services, finance, and other critical applications that require accurate and reliable information to ensure informed decisions.
Data Security Model Using (AES-LEA) Algorithms for WoT Environment
Full text linkThe Web of Things WoT connects physical objects and displays them in the WWW. The growing number of Internet of Things IoT devices and data sharing has led to the attack on sensitive information and allowing unauthorized persons to access and manipulate it. Therefore, ensuring data privacy and protection is a major challenge for organizations and individuals. This paper presents a new hybrid method for the encryption of information to be more suitable for embedded devices in the WoT environment by modifying the Advanced Encryption Standard AES algorithm and hybridizing it with the Lightweight Encryption Algorithm LEA algorithm as well as the Secure Hash Algorithm version 3 SHA3-256 algorithm for integrity and using four dimensional-NSJR system for generation of chaos keys. The proposed method is designed to decrease encryption /decryption time for the information transmitted in the WoT environment and in areas such as government data that need protection against attacks. The proposed method comprises three sub-layers: Chaos Keys generation layer, Data encryption layer, and Authentication layer. The proposed method passed all 15 NIST ( National Institute of Standards & Technology) tests. The amount of time needed to encrypt and decrypt the proposed method was compared with the original encryption methods for different data sizes and five sensitivity levels, and the proposed encryption method was found to be up to (150%) faster while maintaining security strength
Land Cover Change Detection in Iraq Using SVM Classification: A Remote Sensing Approach
Full text linkLand Cover and Land Use studies play an important role in regional socioeconomic development and natural resource management. They support sustainable development by tracking changes in vegetation, freshwater quantity and quality, land resources, and coastal areas. Iraq\u27s Land Use and Land Cover Monitoring with Remote Sensing Data in the Period 2019–2023. This paper performed land use/land cover LULC type classification and time series analysis using Sentinel-2 satellite imagery for the years 2019 and 2023 to identify changes over time. Remote sensing data is used in this paper to address the challenge of detecting land cover change in Iraq through SVM classification. This goal aims to develop a fundamental method of mapping and monitoring these changes, encouraging sustainable land use practices, and achieving the United Nations Sustainable Development Goals. Land cover classes were categorized into five main types: Water, Barren, Building, Vegetation, and Rangeland. The study showed a marked increase in urbanization, and most of this occurring in previously bare soils at the edges of cities. This urbanization was primarily driven by population growth and economic development. What is beneficial for the environment can also be beneficial for us as people humanity as these findings have major implications for urban planning, green space management, and sustainable city development. It seems that there was no change to the existing barren land and buildings, which increased by 8% and 11% respectively, as noted from the data up to October 2023. However, vegetation coverage decreased by 27%, indicating a significant loss of green area. The water category was also up 9%. Results showed satisfactory accuracy assessment (OA: 93.11%) from applying a Support Vector Machine SVM for the LULC classification. The study lays the foundation for ongoing monitoring of LULC changes in Iraq
Fog Computing and IoT Integration for Latency Reduction: A Case Study
No full textشهد إنترنت الأشياء نموًا سريعًا وواسع النطاق، مما أدت الحاجة إلى هياكل حوسبة قادرة على العمل في الزمن الحقيقي وبزمن استجابة منخفض. في الحالات الاعتياديةً هناك تأخير كبير في أنظمة الحوسبة السحابية التقليدية نتيجة للموقع الجغرافي وازدحام الشبكة، مما يجعل هذه الأنظمة غير ملائمة في الحالات التي يكون فيها الزمن عاملًا حاسمًا. في هذه البحث، تم مقارنة أداء أربعة تصاميم معمارية من حيث زمن الاستجابة، وهي: السحابة فقط، الحافة فقط، الضباب متعدد الطبقات، والهجين الذكي المرن.
تم تنفيذ محاكاة واسعة النطاق لتمثيل بيئة إنترنت الأشياء في مدينة ذكية تحتوي على 1000 نوع مختلف من الأجهزة، مثل المستشعرات الذكية، الكاميرات، الأجهزة القابلة للارتداء، وغيرها، والتي تولد حركة بيانات ديناميكية ومتغيرة باستمرار. اظهرت النتائج أن بنية "الحافة فقط" سجلت أقل متوسط زمن استجابة بمقدار 0.2417 ثانية، وتليها البنية الهجينة بزمن قدره 0.3156 ثانية، ثم بنية الضباب متعدد الطبقات عند 0.3874 ثانية. ومن ناحية أخرى، سجل نموذج "السحابة فقط" أطول زمن استجابة بلغ 0.5231 ثانية، مما يشير إلى أن الاعتماد على المعالجة المركزية فقط لا يعد الخيار الأمثل عند التعامل مع التطبيقات الحساسة للزمن.
وقد أدى نموذج "الهجين الذكي"، المبني على آلية التكيف الفوري لعرض البيانات، إلى تحسينات كبيرة بفضل مرونة توزيع المهام بين بيئات الحوسبة السحابية والحوسبة على الحافة. وأظهر تحليل التباين أحادي الاتجاه وجود فروق ذات دلالة إحصائية كبيرة بين التكوينات المختلفة (p < 0.0001). وتبرز أهمية هذه النتائج في دعم التوجه نحو تطوير بنى هجينة ذكية قادرة على التكيف مع السياق لتقليل زمن الاستجابة عند تنفيذ تطبيقات المدن الذكية. كما أن إطار المحاكاة المقترح يُعد أداة قابلة للتوسع وعملية لتوجيه تصميم أنظمة إنترنت الأشياء المستقبلية التي تتطلب استجابة فورية.The Internet of Things (IoT) has risen rapidly and created a need for computing architectures with the ability to aid real-time applications with the least latency. Conventional cloud-only setups often experience high latency due to physical distance and network congestion, thus, unsuitable for time-sensitive applications. Four of the architectural models currently proposed in the literature are assessed in the present paper: Cloud-only, Edge-only, Fog Multi-layer, and Smart Hybrid; each of them is evaluated in a large-scale simulation aimed at simulating a smart city environment with 1,000 various IoT devices, such as sensors, cameras, and wearables generating dynamic traffic. The outcomes indicate that Edge-only architecture recorded the least mean latency of 0.2417 seconds compared to Smart Hybrid with 0.3156 seconds and Fog Multi-layer with 0.3874 seconds mean latencies whereas the Cloud-only model presented the highest latency with 0.5231 seconds. Given that there existed significant differences between the architectures (p < 0.0001) based on the results of the statistical analysis carried using the one-way ANOVA. Such results explain the promise of adaptive and context-sensitive hybrid systems and allow reducing latency in any IoT deployment in the future and designing an IoT-oriented system that supports latency-sensitive applications in smart cities in a more scalable manner
IoT intrusion detection system based on machine learning and deep learning
Full text linkThe proliferation of Internet of Things IoT devices has amplified cybersecurity challenges, necessitating robust Intrusion Detection Systems IDS to safeguard against threats such as botnets and Distributed Denial-of-Service DDoS attacks. This paper evaluates the performance of Machine Learning ML and Deep Learning DL models on two benchmark datasets, BoT-IoT and CIC-IDS2017, to develop efficient IDS. Among ML models, XGBoost demonstrated the best performance, achieving 99.99% accuracy on BoT-IoT and 99.91% on CIC-IDS2017 with superior computational efficiency. For DL, Convolutional Neural Networks CNNs achieved 99.99% accuracy on BoT-IoT and 99.61% on CIC-IDS2017 with preprocessing, highlighting the critical role of data preparation. These findings underline the effectiveness of advanced ML/DL models and preprocessing techniques in enhancing IoT security, providing a pathway for real-time, scalable intrusion detection in IoT environments
Hybrid Optimized Feature Selection and Deep Learning Method for Emotion Recognition That Uses EEG Data
Full text linkخلاصة
المقدمة: يمثل هذا البحث خطوة مهمة نحو تحسين التفاعل بين الإنسان والآلة. يهدف هذا البحث إلى استغلال إمكانيات تخطيط أمواج الدماغ (EEGs) في التعرف على العواطف، وهي مهمة معقدة ومتغيرة. يقدم هذا البحث إطار عمل متكامل لتعزيز التعرف على العواطف، مما يوفر طريقة بديهية للتفاعل العاطفي بين الإنسان والآلة من خلال فهم الآلة للمشاعر البشرية.
المنهجية: تبدأ العملية بجمع ومعالجة بيانات EEG لاستخدامها في تدريب النظام واختباره. تم تطبيق خوارزميات التحسين والتعلم الآلي والتعلم العميق. أولاً، تُستخدم خوارزمية تحسين سرب الجسيمات (PSO) لتحديد وتحسين الوظائف الحرجة وتقليل الأبعاد. بعد ذلك، يتم تطبيق الشبكات العصبية التكرارية (LSTM، GRU، RNN) لتحديد العواطف. تم تقييم جميع النماذج باستخدام مقاييس التقييم الشائعة مثل الدقة، والدقة الإيجابية (precision)، ومقياس F1.
النتائج: من خلال تطبيق النماذج المختلفة لتحديد العواطف باستخدام بيانات EEG، حقق نموذج LSTM نتائج جيدة، حيث بلغت دقته 98.13%، والدقة الإيجابية 98.15%، ومقياس F1 بنسبة 98.13%. بالرغم من أن نماذج GRU وRNN قدمت أداءً جيداً في تحديد العواطف، إلا أن أداء LSTM كان الأفضل.
الأصالة: دمجت الدراسة مفاهيم خوارزمية PSO لاختيار الميزات ونموذج التعلم العميق باستخدام LSTM لتعزيز تحديد العواطف باستخدام EEG. يتغلب النموذج المقترح على الصعوبات المرتبطة بإشارات EEG، مما أدى إلى نظام دقيق لتحديد العواطف، وزاد من فهم الآلة للتفاعل بين الإنسان والآلة. تظهر النتائج الواعدة لإستخدام التعلم العميق في هذا المجال إمكانية التقدم الكبير لتحقيق فهم أعمق للعواطف البشرية من قبل الآلة.Introduction: This study represents an important development in human–machine interactions. It aims to utilize the potential of electroencephalograms (EEGs) in recognizing emotions, which is a complex and variable task. This study presents a complete framework for enhancing emotion identification. It provides an intuitive way for humans to interact with machines emotionally by understanding the emotion machine. The process begins with collecting and preprocessing EEG information to use the data for training and testing the proposed system. Optimization, machine learning, and deep learning algorithms are applied in this study. First, particle swarm optimization (PSO) identifies and optimizes critical functions and reduces feature dimensionality. Thereafter, long short-term memory (LSTM), gated recurrent unit (GRU), and simple recurrent neural network (RNN) architectures are used in emotion identification. All the applied models are evaluated using common evaluation metrics, such as accuracy, precision, and the F1 score. From various implementations of the different models applied to identify EEG emotion, the LSTM model achieved good results with an accuracy of 98.13%, a precision of 98.15%, and an F1 score of 98.13%. Although the GRU and simple RNN models exhibit good performance in emotion identification, their measurements are less than those of LSTM, which outperforms all the other models. This study incorporates the concepts of the PSO algorithm into a feature selection and deep learning model by using LSTM to enhance EEG emotion identification. The proposed model overcomes difficulties and issues related to EEG signals, leading to an accurate emotion detection system and providing enhanced machine understanding of human–machine interactions