Hate speech on online platforms, characterized by discriminatory language targeting individuals or groups, poses significant harm and necessitates robust detection methods for digital safety. Recognizing the ease with which individuals can engage in such speech online, our study delved into detecting Turkish hate speech using deep learning algorithms and natural language processing techniques. We developed innovative methodologies, including a k-means+textGCN classifier with BERT, which marked the first such attempt in the literature, and explored multiple vector representation techniques such as Term Frequency, Word2Vec, Doc2Vec, and GloVe. Additionally, we investigated various learning algorithms and natural language processing techniques, conducting thorough evaluations on three distinct Turkish hate speech datasets. Notably, our newly presented algorithm exhibited superior performance, achieving an impressive F1-score of 87.81% on the 9K dataset, showcasing advancements in hate speech detection and contributing to a safer online environment.

Background/Objectives: The integration of deep learning and computer vision into healthcare has improved medical diagnosis and image analysis. Among object detection algorithms, the YOLO family has attracted substantial attention due to its ability to analyze images in real time with reported improvements in detection performance across multiple studies. This systematic review examines the evolution of YOLO algorithms for diagnostic applications in healthcare from YOLOv1 to YOLOv12. Methods: Peer-reviewed scientific articles published up to 1 January 2026 were retrieved from major scientific databases in accordance with PRISMA 2020 guidelines. The included studies applied YOLO models to medical imaging tasks, including disease and lesion detection and support for clinical procedures. Performance was synthesized using reported metrics such as average precision, accuracy, inference time, and computational efficiency. Results: The reviewed literature suggests progressive architectural refinements associated with reported improvements in diagnostic performance. YOLOv5 and YOLOv8 are the most frequently used architectures in diagnostic settings, reflecting a favorable trade-off between accuracy and computational complexity. YOLO-based methods have demonstrated strong performance across radiological, pathological, ophthalmological, and endoscopic applications. Conclusions: YOLO models have matured into robust and optimized solutions for medical image analysis; however, challenges remain in interpretability, cross-institution generalization, and deployment on edge devices. Future work on explainable YOLO-based diagnostics and energy-efficient model design will be particularly valuable.

Son yıllardaki gelişen teknolojiler neticesinde her türlü hesaplama cihazının İnternete bağlanması sağlanmıştır. Bu sayede birçok gerçek dünya problemi yeni ağ düzenine aktarılsa da bu tam-kontrol sağlanamayan sanal platform çok sayıda güvenlik açığı içermektedir. Günümüzde ağ yöneticilerin ana görevlerinden biride bu açıkları kapatmak ve sorumlu oldukları bilgisayar ağını saldırılardan korumaktır. Güvenlik duvarlarının kullanımı dışarıdan yapılan saldırıları ciddi boyutta engellese de içeriden yapılabilecek veya daha önceden karşılaşılmayan tipten saldırılara karşı zafiyetler içermektedir. Saldırı Tespit Sistemleri (STS) bu zafiyetleri ortadan kaldırmak için öncelikle tercih edilebilecek uygulamalardır. Son geliştirilen STSleri incelendiğinde dinamik bir güvenlik mekanizması geliştirmek adına özellikle Makine Öğrenmesi tabanlı sistemlere ağırlık verildiği görülmektedir. Bilgisayar donanımları ve paralel hesaplama teknolojilerinde ortaya çıkan gelişmeler ve Büyük Veri işleme teknolojilerinin, Makine Öğrenmesi tabanlı sistemlerle uyumlu kullanıldığı görülmektedir. Bu çalışmada yedi farklı makine öğrenimi algoritmaları kullanarak STSlerin geliştirilmesi amaçlanmıştır. Elde edilen sonuçlar başarım, eğitim süreleri ve çalıştırma süreleri açısından karşılaştırılarak farklı kriterlere göre uygun algoritmanın ortaya konmuştur. Bu karşılaştırma için genel kabul gören NSL-KDD veri setinden faydalanılmıştır. Başarım sonuçlarına bakınca Adaboost algoritmasının en yüksek doğruluk oranına ulaştığı görülmektedir. Ancak gerek eğitim süresi gerekse çalışma zamanı performansı göz önüne alınınca Karar Ağacı algoritmasının daha yüksek performans gösterdiği, doğruluk oranı değeri itibarı ile de Adaboost’a yakın değere sahip olduğu görülmektedir.

Machine learning models used for distributed architectures consisting of servers and clients require large amounts of data to achieve high accuracy. Data obtained from clients are collected on a central server for model training. However, storing data on a central server raises concerns about security and privacy. To address this issue, a federated learning architecture has been proposed. In federated learning, each client trains a local model using its own data. The trained models are periodically transmitted to the central server. The server then combines the received models using federated aggregation algorithms to obtain a global model. This global model is distributed back to the clients, and the process continues in a cyclical manner. Although preventing data from leaving the clients enhances security, certain concerns still remain. Attackers can perform inference attacks on the obtained models to approximate the training dataset, potentially causing data leakage. In this study, differential privacy was applied to address the aforementioned security vulnerability, and a performance analysis was conducted. The Data-Unaware Classification Based on Association (duCBA) algorithm was used as the federated aggregation method. Differential privacy was implemented on the data using the Randomized Response technique, and the trade-off between security and performance was examined under different epsilon values. As the epsilon value decreased, the model accuracy declined, and class prediction imbalances were observed. This indicates that higher levels of privacy do not always lead to practical outcomes and that the balance between security and performance must be carefully considered.

Misinformation on social media platforms poses significant risks to public opinion and decision-making due to its rapid dissemination. Traditional text-based approaches often struggle to capture the complex interactions between textual content, user behavior, and contextual relationships. To address this limitation, we propose a heterogeneous graph-based misinformation detection framework integrating user, tweet, and hashtag relations with diverse text representations. In our design, tweets are encoded using both sparse lexical features (Term Frequency–Inverse Document Frequency (TF-IDF)) and dense semantic embeddings (Sentence-Bidirectional Encoder Representations from Transformers (SBERT)), while classification is performed through multiple Graph Neural Network (GNN) architectures, namely Graph Sample and Aggregate (GraphSAGE), Graph Convolutional Network (GCN), and Graph Attention Network (GAT). To account for potential class imbalance, class-weighted loss functions are incorporated during training. Experiments on the TruthSeeker dataset demonstrate that GraphSAGE consistently outperforms GCN and GAT across all feature settings. In particular, the best-performing configuration—TF-IDF combined with GraphSAGE—achieves 94% Accuracy and 94% F1-score, with comparable performance under both standard and class-weighted training. Analysis of training dynamics indicates stable convergence across configurations; GraphSAGE-based models continue to benefit from longer training, reaching 96.73% accuracy at 200 epochs. Comparative evaluations suggest that modeling heterogeneous relations with diverse feature representations provides advantages over purely text-based baselines within the examined setting. These results highlight the value of integrating relational graph structures with complementary feature representations for misinformation detection. While class-weighted training does not substantially alter performance under the near-balanced dataset distribution, it maintains stable learning across classes without degrading overall results. The proposed framework provides a foundation for future extensions, including federated learning, explainability techniques, and cross-domain applications in multilingual contexts.