https://repository.sustech.edu/handle/123456789/9171 2026-09-27T18:43:13Z https://repository.sustech.edu/handle/123456789/28511 Energy Theft Detection in Smart Grids Using Anomaly Detection Techniques with the UCI Smart Meter Dataset Elkhalifa, Tasneem Eltayeb Mohamed; Elzubair, Maram Jamal Eltahir; Tag Elsir, Rawya Elnazir; Taj Eldin, Ruba Mohammed Elamin The study aimed to evaluate the effectiveness of Anomaly Detection techniques in detecting electrical power theft within Smart Grid environments by comparing the performance of Isolation Forest and Autoencoder models using the UCI Smart Meter Dataset. The study was based on an experimental approach, where the data underwent Data Preprocessing processes that included cleaning the data, processing missing values, normalizing the data, extracting properties, and then dividing it into training and testing data. The two unsupervised models were then trained and their performance evaluated using Accuracy, Precision, Recall, F1-Score, Specificity, and AUC-ROC indices, as well as ROC and Precision–Recall curves. The results showed that the Isolation Forest model outperformed the Autoencoder model across all performance indicators, achieving Accuracy of 96.8%, Precision of 83.1%, Recall of 81.9%, F1-Score of 82.5%, Specificity of 98.3%, and AUC-ROC of 99.0%, indicating its stronger performance in identifying abnormal consumption patterns associated with potential energy theft and reducing false alarm rates. The study recommends adopting the Isolation Forest model in Smart Grid monitoring systems, using more diverse datasets to improve the models' generalization ability, developing hybrid models to increase detection efficiency, and testing the models in real-world operating environments that support real-time monitoring. 2026-09-01T00:00:00Z https://repository.sustech.edu/handle/123456789/28510 Intrusion Detection in Mobile Networks Using Random Forest and CICIDS2017 Dataset Abdalla, Safia Abdelrazeg Mohamed; Alemam, Ahmed Noor Alhuda Alzain; Suliman, Alaa Hassan Ahmed; Supervisor;, Rashid A SAEED The rapid growth of mobile networks and the increasing number of cyberattacks have created significant security challenges, making effective intrusion detection essential. To address these challenges, this project proposes a Machine Learning-Based Intrusion Detection System (IDS) for mobile networks using the CICIDS2017 dataset and the Random Forest algorithm. The CICIDS2017 dataset contains realistic network traffic, including both benign and malicious activities with different types of attacks. The proposed system preprocesses the dataset by removing missing and invalid data, handling outliers, encoding labels, and scaling the features. Recursive Feature Elimination (RFE) is applied to select the most relevant features before training the Random Forest model. The model is trained to distinguish between normal and malicious network traffic and to classify different attack types. The performance of the proposed system is evaluated using Accuracy, Precision, Recall, F1-Score, Specificity, and AUC-ROC. The results demonstrate that Random Forest can effectively detect and classify network attacks with high performance. The proposed system can therefore contribute to improving the security, reliability, and protection of mobile network environments against cyber threats. 2026-09-01T00:00:00Z https://repository.sustech.edu/handle/123456789/28509 An Enhanced Intelligent Server Monitoring System for Predictive Fault Detection and Triage Prioritization Ibrahim, Ahmed Abdulmonim Mohamed; Alnour, Alrahma Fadol Allah Gomaa; Mohamed, Marwa Yousuf; Abdelsalam, Maria Adil Mohamed; Ali, Abdalmajed Mohammed Abdalrheem Modern digital infrastructure relies heavily on continuous server availability, yet traditional monitoring tools remain largely reactive, relying on static thresholds that generate excessive false alerts and often miss early signs of degradation. This research presents an enhanced intelligent server monitoring system that combines real-time stream processing with a hybrid unsupervised machine learning model for predictive fault detection and triage prioritization. The system ingests server metrics (CPU, memory, disk, and network utilization) through an Apache Kafka message bus and computes a weighted hybrid anomaly score. S(x_t) that combines an Isolation Forest outlier score with an Autoencoder reconstruction error, and classifies detected anomalies into priority levels using a Random Forest classifier. Prometheus and InfluxDB provide short-term and long-term time-series storage, respectively; Grafana delivers live visualization dashboards; Telegram delivers real-time alerts. An independent Apache Spark batch layer periodically generates aggregate historical reports. The models were trained on 2,243 real records from the Alibaba Cluster Trace 2018 production dataset, selected after evaluating and rejecting two alternative public datasets (the Server Machine Dataset, whose 38 metrics are anonymized, and the NAB AWS CloudWatch collection, which lacks a memory-utilization metric). The hybrid score's decision threshold (0.4963) was derived automatically from the 98th percentile of the training score distribution. System performance was validated across 28 automated tests spanning unit, integration, and deployment levels (100% pass rate), and the anomaly detector was evaluated against a test set with known injected anomalies, achieving a 92.5% detection rate (Recall) with a 3.39% false-positive rate. The results demonstrate that a hybrid unsupervised approach, trained on real production telemetry, can deliver practical, low-noise anomaly detection suitable for real-time server health monitoring. 2026-09-01T00:00:00Z https://repository.sustech.edu/handle/123456789/28508 Adaptive Federated Learning-Based Resource Management in Mobile Edge Computing for Machine Type Communication in 5G Network Ahmed, Asia Easa Yousuf; Hassan, Esraa Yahya Zakaria; Ali, Omnia Ahmed Abd Elgadir; Mohamed, Ruaa Saadeldin Abdelhameed; Hassan, Yusra Yahya Zakaria With the rapid expansion of intelligent devices and the growing demand for real-time data processing, massive volumes of data are now generated and stored at the network edge rather than in centralized clouds. To ensure data privacy while enabling collaborative model training, Federated Learning (FL) has emerged as a promising paradigm that allows devices to train local models using private data and share only model parameters with a central server. However, due to limited wireless bandwidth and constrained energy resources in mobile devices, it is impractical for FL to perform model updates and aggregation across all participating devices simultaneously. Furthermore, selecting appropriate devices for each training round is a critical challenge that directly impacts resource utilization, energy efficiency, and communication latency. In 5G-enabled Mobile Edge Computing (MEC) environments supporting Machine Type Communication (MTC), these challenges are further amplified by device heterogeneity and dynamic network conditions, making efficient coordination and resource management essential for maintaining performance and scalability. To address these issues, we propose an adaptive resource management framework for MEC-assisted FL that integrates intelligent device selection and dynamic resource allocation using a Double Deep Q-Network (DDQN). The device selection process is modeled as a Markov Decision Process (MDP), allowing the DDQN to make dynamic and experience-driven decisions that balance energy consumption, computational load, and bandwidth utilization under varying network and device conditions. By combining 5th generation network (5GN) capabilities with FL on the MNIST dataset, the framework achieves an intelligent, low-latency, and energy-efficient edge learning environment—paving the way for scalable, privacy-preserving, and real-time distributed Artificial Intelligence (AI) systems for MTC applications. 2026-09-01T00:00:00Z