Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed
Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741
Browse
5 results
Search Results
Publication Metadata only Fault Estimation for Operational Systems(Institute of Electrical and Electronics Engineers Inc., 2022) Ozkent, Tuncberk; Gelal, Ece; Ozkent, Tuncberk, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gelal, Ece, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyOperational systems are crucial for corporations. A majority of the business processes flow through these systems and even minor downtimes on these systems may cause serious financial consequences. System logs are a promising way for analyzing the behaviors of operational systems. This work investigates fault estimation on a real-life data set derived from the system logs of a large-scale insurance company. Data set consists of operational system indicators like visit duration, connection properties and time of connection collected over four months. Regression and classification algorithms have been used to estimate the impact of the system and environmental parameters on the system response time. The best performance is obtained with the CatBoost classification, which yields 99% accuracy in estimating whether system responds within normal interval. This study assists the operational team in identifying problem scenarios, future improvements may be possible as logs from other operational systems from the company are considered using transfer learning. © 2022 Elsevier B.V., All rights reserved.Publication Metadata only QoE Estimation for the Wi-Fi Edge with Gradient Boosting-based Machine Learning(Institute of Electrical and Electronics Engineers Inc., 2023) Argın, Berke; Demir, Mehmet Özgün; Onalan, Aysun Gurur; Salik, Elif DIlek; Gelal, Ece; Argın, Berke, Lifemote Networks, Istanbul, Turkey; Demir, Mehmet Özgün, Lifemote Networks, Istanbul, Turkey; Onalan, Aysun Gurur, Lifemote Networks, Istanbul, Turkey; Salik, Elif DIlek, Lifemote Networks, Istanbul, Turkey; Gelal, Ece, Bahçeşehir Üniversitesi, Istanbul, TurkeyAn integral part of the Intent-Based Networking paradigm is estimating and improving the end-user quality of experience (QoE). Estimating user experience from the (wide-area) network data alone does not accurately represent the performance at customer premises since Wi-Fi at the edge also significantly affects the perceived QoE. We propose machine learning-based estimation of the end-users' perceived QoE for web browsing and video streaming applications, based on Wi-Fi statistics. We implement support vector machine (SVM), decision tree (DT), multilayer perceptron (MLP), XGBoost, and CatBoost algorithms and compare their performance. To the best of our knowledge, our CatBoost-based model yields the highest accuracy to date, 0.92 R2, in estimating the QoE for web browsing based on Wi-Fi statistics. Our experiments also show that the XGBoost-based QoE estimator outperformed the neural network-based model in estimating the QoE for video streaming. Our work demonstrates that network operators can infer the user-perceived QoE in a Wi-Fi network through telemetry data obtained by passive measurements. © 2023 Elsevier B.V., All rights reserved.Publication Metadata only Unmasking Twitter Bots: Feature Engineering and Machine Learning for Bot Account Identification(Institute of Electrical and Electronics Engineers Inc., 2023) Babur, Zeynep; Bekdemir, Umut; Sen, Acelyanur; Carkit, Sevval Ozlem; Genc, Oguzhan; Gülcü, Ayla; Gümüştaş, Cihangir; Gelal, Ece; Babur, Zeynep, Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Bekdemir, Umut, Department of Software Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Sen, Acelyanur, Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Carkit, Sevval Ozlem, Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Genc, Oguzhan, Department of Software Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gülcü, Ayla, Department of Software Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gümüştaş, Cihangir, Department of Engineering Management, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gelal, Ece, Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyBot accounts on X (formerly known Twitter)1 are a growing issue that limits and negatively impacts the browsing and sharing experience of Twitter users, and it is important to identify such accounts. In this work, we perform machine learning-based estimation of bot accounts on Twitter. Using Twitter's API, a data set is collected containing tweets and related metadata from various accounts. Feature engineering techniques are then applied to highlight relevant features such as sentiment analysis of the account's tweets, or the account's friend/follower ratio. Using these features to train and evaluate machine learning models, the likelihood of a given account being a bot is estimated. The performances of three different models are comparatively analyzed based on their fl score, accuracy, precision, and recall. Analysis of feature importances shows the success of derived features in identifying bot accounts. This work demonstrates the potential of using feature engineering with tweet and profile properties to detect bot accounts on Twitter, and provides a foundation for further research on this topic.1In October 2022, Twitter was acquired, after which Twitter Inc. ceased to exist as an independent company and was merged with X Corp. To prevent possible inconsistencies and confusion in literature search results, we used the name Twitter throughout the text. © 2023 Elsevier B.V., All rights reserved.Publication Metadata only Advancing WebRTC QoE Assessment with Machine Learning in Real-World Wi-Fi Scenarios(Institute of Electrical and Electronics Engineers Inc., 2024) Argın, Berke; Demir, Mehmet Özgün; Salik, Elif DIlek; Onalan, Aysun Gurur; Batum, Öykü Han; Gelal, Ece; Argın, Berke, Lifemote Networks, Istanbul, Turkey; Demir, Mehmet Özgün, Lifemote Networks, Istanbul, Turkey; Salik, Elif DIlek, Lifemote Networks, Istanbul, Turkey; Onalan, Aysun Gurur, Lifemote Networks, Istanbul, Turkey; Batum, Öykü Han, Lifemote Networks, Istanbul, Turkey; Gelal, Ece, Bahçeşehir Üniversitesi, Istanbul, TurkeyVideo conferencing applications play a key role in enabling use cases like remote working, education, and potentially the metaverse. From the perspective of Internet service providers, predicting the end user’s Quality of Experience (QoE) in such applications is critical in allocating the right resources to ensure consistently high QoE. This work addresses the estimation of user QoE from link-layer performance metrics such as transferred packets, queue size, signal strength, and channel occupancy for WebRTC-supported applications. Our study entails collecting a data set capturing various Wi-Fi scenarios in practical environments and training machine learning models on this data to estimate the perceived QoE. Our findings demonstrate improvement in prediction accuracy compared to earlier models and QoE representations, furthermore, we also investigate the explainability of the models with the help of SHAP values. © 2024 Elsevier B.V., All rights reserved.Publication Metadata only Analysis of Adversarial Training for Resilient Image Recognition Against Different Attacks(Institute of Electrical and Electronics Engineers Inc., 2024) Karakuş, Emir Can; Islim, Firat Kutluhan; Gelal, Ece; Adali, E.; Karakuş, Emir Can, Bahçeşehir Üniversitesi, Istanbul, Turkey; Islim, Firat Kutluhan, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gelal, Ece, Bahçeşehir Üniversitesi, Istanbul, TurkeyImage recognition systems are used for various critical use cases, from cancer detection to autonomous vehicles, while their accuracy closely relies on the data that they are trained on. Recently, adversarial machine learning has been flagged a s a possible threat against the successful operation of such systems and models. While the research towards understanding how to manipulate such models continues, one mitigation approach is to increase the model's resilience against such samples via including such samples in the training process. In this work, we examine the impact of adversarial training on improving robustness against other adversarial attack types. We generate adversarial samples via two image perturbation methods, Fast Gradient Sign Method (FGSM) and Jacobian-based Saliency Map Attack (JSMA), and include these samples respectively in the training set of two independent ResNet-18 models on the CIFAR-10 dataset. Obtaining two adversarially-trained models, we compare their accuracies upon each attack and discuss the impact on model resiliency when the training set includes samples generated by different approaches. Our results highlight that FGSM-fine tuned (adversarially trained) model earns the model greater resilience against FGSM attack compared to JSMA attack. Adversarial JSMA-training, when JSMA samples are generated to target the original class, earns the model resilience against both attacks. © 2025 Elsevier B.V., All rights reserved.