https://hdl.handle.net/20.500.12619/842 2026-04-11T18:00:35Z 2026-04-11T18:00:35Z Buken, Ayse Betul Akgun, Devrim https://hdl.handle.net/20.500.12619/103351 2026-03-25T14:23:32Z 2025-06-01T00:00:00Z

Anomaly Detection in IoT Device Traffic Data Using Deep Learning Buken, Ayse Betul; Akgun, Devrim The Internet of Medical Things (IoMT) is important in modern healthcare systems by enabling continuous monitoring and data­driven medical services. However, adopting IoMT is associated with cybersecurity challenges due to the different protocols and vulnerabilities to network­based threats. This study examines anomaly detection in IoMT traffic utilizing the CICIoMT2024 dataset, a comprehensive benchmark that includes 18 cyberattacks across 40 devices in Wi­Fi, Bluetooth, and MQTT environments. We evaluate machine learning models such as Logistic Regression, Random Forest, AdaBoost, and Deep Neural Networks (DNN). Additionally, we explore deep learning architectures that combine convolutional layers with Long Short­Term Memory (LSTM) units. Our experiments cover a range of model assessments for binary, six­class, and nineteen­class classification tasks. Model performance is measured using standard metrics such as accuracy, precision, recall, and F1­score. This research demonstrates the potential of deep learning to enhance the security of IoMT networks and supports further development of better detection models in the healthcare environment.

2025-06-01T00:00:00Z ÇAĞLAR, Betül Sena Akgün, Devrim https://hdl.handle.net/20.500.12619/103020 2025-05-07T10:57:11Z 2024-12-25T00:00:00Z

ANOMALY DETECTION IN INDUSTRIAL MACHINES USING EXPLAINABLE AI AND ACOUSTIC SIGNALS ÇAĞLAR, Betül Sena; Akgün, Devrim In recent years, anomaly detection in industrial machinery has become a critical topic aimed at preventing failures, improving maintenance processes, and enhancing production efficiency. As industrial machines operate continuously under varying conditions, early detection of anomalies can reduce unplanned delays and maintenance costs. Acoustic signals are important data sources for evaluating the health of industrial equipment. This study proposes a lightweight machine learning (ML) approach for anomaly detection in industrial machines using acoustic signals. For this purpose, audio features are extracted, and their statistical metrics are calculated to define a new dataset for anomaly detection. Explainable Artificial Intelligence (XAI) techniques are employed to determine the most effective features that contribute to anomaly detection of machines. The proposed method tested using the publicly available MIMII Dataset: Sound Dataset for Malfunctioning Industrial Machine Investigation and Inspection. The results demonstrate that the proposed approach achieves highly promising outcomes, with an accuracy of 99.7% and an AUC score of 99.9% in classifying anomalies in industrial machines.

2024-12-25T00:00:00Z Akgun, Devrim https://hdl.handle.net/20.500.12619/101337 2024-01-24T07:05:10Z 2023-12-25T00:00:00Z

Working Principles of Convolutional Neural Networks in Keras Akgun, Devrim

2023-12-25T00:00:00Z Alyafawi, Z. , Akgun, D. https://hdl.handle.net/20.500.12619/101336 2023-12-26T05:54:14Z 2022-12-29T00:00:00Z

Speech Recognition Using Deep Learning Model With Volterra Series-Based Layers in Tensorflow Alyafawi, Z. , Akgun, D. The Volterra series is a mathematical tool widely used to analyze and model nonlinear systems. The Volterra model expands a nonlinear system's response in terms of a series of integral equations. Like linear convolution, nonlinear convolution operators can be integrated into deep learning layers. This research proposes a new layer based on a second-order 1D Volterra series expansion using the TensorFlow environment. To develop the Volt1D, we first analyzed a linear convolutional layer's performance on a human speech dataset. The Volterra series has been particularly successful in speech recognition, as it allows for modeling the nonlinear dynamics of the human vocal tract. Volt1D allowed us to capture higher-order nonlinearities in the system, significantly improving the model's accuracy. To validate the effectiveness of the Volt1D, we conducted extensive experiments on a dataset of the human speech command. Overall, our research demonstrates the potential of the Volt1D as a powerful tool for training speech recognition models.

2022-12-29T00:00:00Z