Hande Çavşi Zaim

Mrs. Hande Çavşi Zaim

Computer Engineering, Eskişehir Osmangazi University, Eskişehir, Turkey (Türkiye)
orcid

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 January 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
HZHande Çavşi Zaim
Hande Çavşi Zaim
EYEsra N. Yolacan
Esra N. Yolacan
The increase in cybersecurity threats has made attack detection systems critically important. Traditional deep learning methods often require large amounts of data and struggle to understand relationships between features effectively. With their self-attention mechanism, Transformers excel in modeling complex relationships and long-term dependencies. They are also adaptable to various data types and sources, making them advantageous in large-scale attack detection scenarios. This paper introduces the FPE–Transformer framework, leveraging the strengths of the Transformer architecture. FPE–Transformer incorporates an innovative feature positional encoding mechanism that encodes the positional information of each feature separately, enabling a deeper understanding of feature relationships and more precise attack detection. Additionally, the model includes a ClassificationHead for enhanced accuracy and complex pattern recognition. The framework’s performance was validated using the NSL-KDD and CIC-IDS2017 datasets, demonstrating its superiority over traditional methods in detecting diverse attack types and improving overall performance. This study highlights FPE–Transformer’s innovative approach and ability to address key limitations of traditional deep learning methods, establishing it as a robust solution for modern attack detection challenges.

27 January 2025 in IEEE Access
Verified
Journal Article
Open access
EYEsra N. Yolacan
Esra N. Yolacan
HZHande Çavşi Zaim
Hande Çavşi Zaim
Robotic systems have become integrated components across various industries, offering improved operational efficiency and automation. However, this wide usage area makes robotic systems susceptible to cyberattacks and security vulnerabilities. Replay and subscriber flood attacks pose significant challenges to the integrity and security of robotic arms. In response to these challenges, this study proposes a new approach that leverages deep learning techniques for attack detection. We present a new framework, LSTMbased Dynamic Compound Weight Mechanism (DCWM), designed to identify cyberattacks targeting robotic arms effectively. In addition, we introduce a new feature reduction method in the preprocessing module. Our methodology is evaluated through ACC and MSE measurements. Compared to simple LSTM models, our proposed framework achieves a significant enhancement in accuracy, with performance gains observed across both replay and subscriber flood attack scenarios. Specifically, the DCWM-LSTM model exhibits an increase in accuracy from 88% to 92% for replay attacks and from 93% to 97% for subscriber flood attacks. These findings underscore the effectiveness and robustness of the DCWM-LSTM framework as a practical solution for detecting and mitigating cyber threats against robotic systems. By enhancing the security posture of robotic arms, our approach contributes to safeguarding critical infrastructure and ensuring the continued reliability and safety of automated processes in diverse industrial settings.

18 November 2023 in Neural Computing and Applications
Verified
Journal Article
Closed access
PDF file
HZHande Çavşi Zaim
Hande Çavşi Zaim
MYMetin Yılmaz
Metin Yılmaz
EYEsra N. Yolacan
Esra N. Yolacan

2 August 2024 in Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi
Verified
Journal Article
Closed access
HZHande Çavşi Zaim
Hande Çavşi Zaim
EYEsra N. Yolacan
Esra N. Yolacan
Sensor fusion involves combining data from multiple sources, resulting in more detailed and reliable results. Sensor fusion techniques play an important role in many application areas for industries such as defense, automotive, military and healthcare. Sensor fusion techniques, which include various approaches, have been examined independently of the sector, and a broader review has been performed. In addition to covering the advantages and challenges of different sensor fusion techniques in-depth, this work also provides a comprehensive classification of sensor fusion algorithms, techniques, and architectures. This comprehensive classification aims to guide researchers in making choices according to their needs by providing an understanding of sensor fusion techniques and algorithms. Additionally, when evaluating the future potential of sensor fusion, the focus is on how fusion techniques may evolve with increasing complexity and diversity. This study provides an in-depth analysis of the existing literature, contributing to the advancement of research in sensor fusion and the development of more effective systems.

29 September 2021 in Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi
Verified
Journal Article
Open access
PDF file
HZHande Çavşi Zaim
Hande Çavşi Zaim
EYEsra N. Yolacan
Esra N. Yolacan
EGEyyup Gulbandilar
Eyyup Gulbandilar
Günümüzde, şirketler gelecekte yapmayı planladıkları işleri içeren çok sayıdaki önemli verilerini elektronik ortamlarda saklamaktadırlar. Saldırı durumunda ise hem şirkete hem de bireylere zarar verebilecek finansal bilgiler hedef alınmaktadır. Bu saldırı türlerinden biri de banka ödemelerinde meydana gelen dolandırıcılık saldırılarıdır. Grafik veri bilimi kullanılması, mevcut analitik ve makine öğrenimi ardışık düzenlerini güçlendirerek, var olan dolandırıcılık tespit yöntemlerinin doğruluğunu ve uygulanabilirliğini arttırmaktadır. Bu çalışmada İspanya’daki bir banka ödeme bilgi simülasyonundan oluşturulan BankSim veri kümesi kullanılmıştır. BankSim üzerinde bulunan normal ödemeler ve sahte veriler sınıflandırılarak dolandırıcılık tespiti gerçekleştirilmesi amaçlanmıştır. Sınıflandırma için Python dilinde RandomForest (RF), Support Vector Machine SVM, XGBoost (XGB), K-Nearest Neighbors (k-NN) sınıflandırma algoritmaları kullanılmıştır. Performans değerlendirmeleri için K-katlamalı çapraz doğrulama kullanılmıştır. Çizge madenciliği için Neo4j veritabanı kullanılmış ve Neo4j sorgu dili olarak CypherQL kullanılmıştır. Bu dolandırıcılık tespitinin uygulanması ile daha az hileli işlem ve daha güvenilir bir gelir akışı elde edilmiştir. Çizge madenciliği aşamasında PageRank, Community, degree gibi çizge algoritmaları ile birlikte standart makine öğrenimi yöntemi ile elde edilen sonuçlar optimize edilmiştir. Bu açıdan çizge madenciliği ve makine öğrenimi algoritmalarının birlikte kullanılmasının diğer yöntemlere kıyasla doğruluk oranlarının daha yüksek olduğu ve daha hızlı sürede hesap yapan bir yöntem olduğu ispatlanmıştır.