BBBüşra Büyüktanir

Büşra Büyüktanir

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Büşra Büyüktanir has a total of 23 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Machine unlearning through model forgetting techniques: application area and emerging methods

    by Busra Buyuktanir, Elif Kucur, Zeynep Kuyumcu, Nese Ozdemir, Sahsene Altinkaya, Gozde Karatas Baydogmus, Kazim Yildiz


  • Evaluating SISA-Based Machine Unlearning Across Diverse Modalities: Tabular, Visual and Auditory Data

    by Arda Kurt, Abdulsamet Cakir, Cemal Can Polat, Büşra Büyüktanır, Gozde Karatas Baydogmus, Kazım Yıldız

    Retraining a model entirely from scratch to delete specific information is highly inefficient in terms of both time and cost. This study comprehensively evaluates the training framework known as Sharded, Isolated, Sliced, and Aggregated (SISA) as a practical solution to this problem. Researchers tested SISA’s unlearning performance across a wide range of data types. These included tabular data from the Purchase and Adult datasets, visual data from the DR Grading dataset, and auditory data from the ESC fifty dataset. The core principle of the method involves partitioning data into small, independent sections. This allows for the retraining of only the relevant component when a deletion request is made. The study’s key findings indicate that SISA’s success is largely dependent on the characteristics of the data. The approach demonstrates exceptional efficiency with large, well-structured datasets, significantly reducing retraining time while maintaining high model accuracy. However, its performance shows fluctuations when applied to more complex or irregular datasets. Ultimately, this work presents the SISA method as a scalable and viable strategy for data privacy in modern artificial intelligence systems, offering a crucial balance between model accuracy, operational efficiency, and legal compliance.

  • Open Access

    Decentralized Disease Prediction: A Federated Learning Perspective

    by Abdülbaki Demir, Ahmet Yasir Kulaksız, Büşra Büyüktanır, Gozde Karatas Baydogmus, Kazım Yıldız

    Machine learning has gained great importance in the healthcare industry due to its ability to extract meaningful information from large datasets. Nonetheless, the sensitive nature of health data presents serious privacy concerns, limiting the feasibility of centralized data collection and the applicability of conventional machine learning algorithms. To address these challenges, this study employs a federated learning-based aggregation algorithm, Data Unaware Classification Based on Association (duCBA), for disease prediction tasks. Model training was conducted using two publicly available health datasets: Diabetes Health Indicators and Heart Attack Prediction. Prior to training, feature selection techniques including Mutual Information, Chi-Square, and Principal Component Analysis (PCA) were applied to generate three distinct datasets. The system architecture comprises a central server, an API server, and multiple clients. Clients locally train models on their own data without sharing it externally. The API server collects these locally trained models and transmits them to the central server. The central server then aggregates them using the duCBA algorithm and disseminates the updated global model back to the clients. This research not only demonstrates the practical viability of duCBA in federated settings but also compares its performance against the widely-used Federated Averaging (FedAvg) algorithm. Experimental results reveal that duCBA achieves comparable performance to FedAvg while offering improved model transparency in line with Explainable Artificial Intelligence (XAI) principles. Notably, the duCBA algorithm delivers stable and consistent accuracy (67–71%) across different feature selection methods and varying client numbers, particularly excelling in scenarios with fewer clients.

  • Open Access

    Differential Privacy in Federated Learning: Mitigating Inference Attacks with Randomized Response

    by Ozer Ozturk, Busra Buyuktanir, Gozde Karatas Baydogmus, Kazim Yildiz

    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.


  • Top Cited Papers
    Open Access

    Federated learning in intrusion detection: advancements, applications, and future directions

    by Busra Buyuktanir, Şahsene Altinkaya, Gozde Karatas Baydogmus, Kazim Yildiz

    Federated Learning (FL) has emerged as a promising distributed machine learning approach that addresses confidentiality and integrity concerns in various sectors, including Internet of Things (IoT), healthcare, finance, and cybersecurity. In order to improve privacy protection and detection accuracy in decentralized systems, this study investigates the incorporation of FL into Intrusion Detection Systems (IDS). FL is especially useful in situations where data security and privacy are crucial because it allows for the cooperative training of models without centralizing sensitive data. We examine many FL-based IDS solutions across several domains, emphasizing how well they mitigate data breaches, maintain confidentiality, and enhance intrusion detection capabilities. The use of Generative Adversarial Networks (GANs), artificial immune systems, and hybrid deep learning techniques to maximize IDS performance are among the current developments in FL methodology that are covered in the paper. We also look at issues like the requirement for effective aggregation procedures and non-independent and identically distributed (non-IID) data. Finally, we outline future directions and open research topics to improve the scalability, resilience, and effectiveness of FL-based IDS solutions in practical applications.


  • Open Access

    Sosyal medyada nefret söylemi çerçevesi: Geçmişten günümüze kapsamlı bir derleme

    by Büşra Büyüktanır, Özlem Yakar, Ayşe Berna Altınel Girgin

    Nefret söylemi, bireye veya gruba yönelik nefret ifadelerini içeren ve her türlü olumsuz uygulama ya da uygulamaya teşvik etmek amacıyla kullanılan ifadelerdir. Bu ifadeler, geçmişten günümüze yayılarak süregelmiştir. Nefret söylemleri kapsamında üretilen olumsuz ifadelerin tümü ve bunların kullanıma yayıldığı ortamların sosyal medya araçları olduğu söylenebilir. Bu araçlar, sosyal açıdan iletişim anlamında toplumsal öneme sahiptir. Teknolojinin gelişimiyle, bireyler geleneksel iletişim yerine dijital iletişimi tercih ettiği için, nefret söylemlerinin sosyal medya ortamlarında yer alması kabul edilebilir bir normdur. Türkçe dilinde nefret söylemlerine sosyal medya ortamlarında sıkça rastlanır; fakat bu konuda yapılan çalışma sayısı oldukça azdır. Ayrıca, sosyal medyada nefret söylemleri günümüzdeki çalışmaları kapsamaktadır. Bu iki düşünceyi bir arada tutan çalışma henüz bulunmamaktadır. Bu çalışma, bahsedilen boşluğu doldurmak amacıyla hazırlanmıştır. Çalışmada, geçmişten günümüze nefret söylemlerinin gelişimi ve Türkçe dilindeki nefret söylemleri detaylı şekilde incelenmiştir. Çalışma kapsamında, literatür incelemesi yapılırken 2019-2023 yıllarındaki çalışmalar dikkate alınmıştır. Çalışmada, konu olarak uygulama tabanlı çalışmalar ve sosyal analiz tabanlı çalışmalar incelenmiştir. Uygulama tabanlı çalışmalarda, nefret söylemi tespiti için makine öğrenimi ve derin öğrenme tabanlı uygulamalara yer verilmiştir. Sosyal analiz çalışmalarında ise, nefret söylemi tespitinde konu temelli dikkat uyandıran çalışmalar, betimsel analiz yapan çalışmalar, söylem analizi yapan çalışmalar ve içerik analizi yapan çalışmalar olarak alt başlıklar halinde incelenmiştir.


  • Open Access

    du-CBA: Veriden habersiz ve artırımlı sınıflandırmaya dayalı birliktelik kuralları çıkarma mimarisi

    by Büşra Büyüktanir, Kazım YILDIZ, Eyüp Emre Ülkü, Tolga Büyüktanir

    İstemci sunucu sistemlerinde makine öğrenmesi modeli kullanılması bir ihtiyaçtır. Ancak istemcilerden verilerin toplanması, sunucuya aktarılması, makine öğrenmesi modeli eğitilmesi ve bu modelin istemcilerde çalışan cihazlara entegre edilmesi bir çok problemi beraberinde getirmektedir. Verilerin istemcilerden sunucuya transferi ağ trafiğine sebep olmakta, fazla enerji gerektirmekte ve veri mahremiyetini istismar edilebilmektedir. Çalışma kapsamında, bahsedilen problemlere çözüm için federe öğrenme mimarisi kullanılmaktadır. Mimariye göre, her bir istemcide istemcinin kendi verilerinden makine öğrenmesi modeli eğitilmektedir. Her bir istemcide eğitilen modeller sunucuya gönderilmekte ve sunucuda bu modeller birleştirilerek yeni bir model oluşturulmaktadır. Oluşturulan nihai model tekrar istemcilere dağıtılmaktadır. Bu çalışmada Veriden Habersiz İlişkili Kurallara Dayalı Sınıflandırma (Data Unaware Classification Based on Association, du-CBA) olarak adlandırılan ilişkisel sınıflandırma algoritması geliştirilmiştir. Federe öğrenme ile klasik öğrenme mimarilerini karşılaştırıp başarılarını ölçmek için çalışma kapsamında benzetim ortamı oluşturulmuştur. Benzetim ortamında du-CBA ve CBA algoritmaları kullanılarak modeller eğitilmiş ve sonuçlar kıyaslanmıştır. Modellerin eğitiminde University of California Irvine (UCI) veri havuzundan alınan beş veri seti kullanılmıştır. Deneysel sonuçlar, her bir veri seti için federe öğrenme ile eğitilen modellerin, klasik öğrenme ile eğitilen modellerle neredeyse aynı doğruluğu elde ettiğini ama eğitim sürelerinin yaklaşık %70 oranında azaldığını göstermiştir. Sonuçlar geliştirilen algoritmanın başarıya ulaştığını ortaya koymaktadır.


  • Open Access

    Raspbraille: Conversion to Braille Alphabet with Optical Character Recognition and Voice Recognition Algorithm

    by Filiz Dalip, Kazım YILDIZ, Eyüp Emre Ülkü, Büşra Büyüktanir

    It is quite difficult for visually impaired individuals to read a physically printed material. It is a necessity to provide information flow and communication between visually impaired individuals and other individuals. For this purpose, an alphabet has been developed for visually impaired individuals. With this alphabet, which is called the Braille alphabet, it is possible for visually impaired individuals to access written culture. There are special letters in the alphabet designed for the visually impaired. With the development of technology, studies in the field of Braille alphabet have increased. Optical character recognition technology can convert the letters on files such as pdf, pictures, photos into understandable data by selecting and separating them. Speech recognition algorithm, computer, tablet, mobile phone, voice recorder devices, etc. It is the process of detecting and recognizing the human voice through a microphone with technological products. Within the scope of the study, a device has been developed for Braille alphabet by using optical character recognition technology and speech recognition algorithm. First, the images of each page of the books that were requested to be translated into Braille alphabet were taken. Then, a transformation was performed using a specific character set with the optical character recognition algorithm. Audiobooks, on the other hand, were first converted into text and then into Braille alphabet using speech recognition algorithm. With these converted characters, a physical hardware device has been developed to display Braille letters that visually impaired individuals can read faster.

  • Housing Price Estimation with Deep Learning: A Case Study of Sakarya Turkey

    by Murat Ozdemi̇r, Kazım YILDIZ, Büşra Büyüktanir

    Shelter is one of the most basic human needs. Besides housing needs, the housing market is also very important for investment. It is also a market where many people, such as engineers, architects, real estate agents make economic gain. When a house is bought for living in it, it is not desired to be changed for many years, and when it is bought for investment, it is a tool that requires good income. Therefore, the best decision should be made when buying a house, and it should be scrutinized. Correct estimation of house prices is very important for both buyers to make the right decision and for sellers to sell without a loss. There are many parameters for estimating house prices. In addition to variables such as the number of floors, location, and several bathrooms used in previous studies, economic factors (such as the price of bread, foreign currency price, new car price) and the housing loan interest rate of the banks were taken as inputs in this study. Sakarya province, where all parameters can be tested to make a more accurate determination, was chosen as the research area. A comparison of polynomial regression, random forest, and deep learning methods was made and it was concluded that the most accurate method was deep learning. At the same time, it was determined which parameters are more effective in house price estimation.