Yazar "Yavuz, Ali Gökhan" seçeneğine göre listele

Listeleniyor 1 - 4 / 4
  • Küçük Resim
    Öğe
    A better way of extracting dominant colors using salient objects with semantic segmentation
    (2021) Gündüz, Ayşe Bilge; Taşkın, Berk; Yavuz, Ali Gökhan; Karsligil, Mine Elif
    One of the most prominent parts of professional design consists of combining the right colors. This combination can affect emotions, psychology, and user experience since each color in the combination has a unique effect on each other. It is a very challenging to determine the combination of colors since there are no universally accepted rules for it. Yet finding the right color combination is crucial when it comes to designing a new product or decorating the interiors of a room. The main motivation of this study is to extract the dominant colors of a salient object from an image even if the objects overlap each other. In this way, it is possible to find frequent and popular color combinations of a specific object. So, first of all, a modified Inception-ResNet architecture was designed semantically segmentate objects in the image. Then, SALGAN was applied to find the salient object in the image since the aim here is to find the dominant colors of the salient object in a given image. After that, the outputs consisted of the SALGAN applied image and segmented image were combined to obtain the corresponding segment for the purpose of finding the salient object on the image. Finally, since we aimed to quantize the pixels of the corresponding segment in the image, we applied k-means clustering which partitions samples into K clusters. The algorithm works iteratively to assign each data point to one of the K groups based on their features. Data points were clustered according to feature similarity. As a result the clustering, the most relevant dominant colors were extracted. Our comprehensive experimental survey has demonstrated the effectiveness of the proposed method.
  • Küçük Resim
    Öğe
    An attack detection framework based on BERT and deep learning
    (IEEE, 2022) Seyyar, Yunus Emre; Yavuz, Ali Gökhan; Ünver, Halil Murat
    Deep Learning (DL) and Natural Language Processing (NLP) techniques are improving and enriching with a rapid pace. Furthermore, we witness that the use of web applications is increasing in almost every direction in parallel with the related technologies. Web applications encompass a wide array of use cases utilizing personal, financial, defense, and political information (e.g., wikileaks incident). Indeed, to access and to manipulate such information are among the primary goals of attackers. Thus, vulnerability of the information targeted by adversaries is a vital problem and if such information is captured then the consequences can be devastating, which can, potentially, become national security risks in the extreme cases. In this study, as a remedy to this problem, we propose a novel model that is capable of distinguishing normal HTTP requests and anomalous HTTP requests. Our model employs NLP techniques, Bidirectional Encoder Representations from Transformers (BERT) model, and DL techniques. Our experimental results reveal that the proposed approach achieves a success rate over 99.98% and an F1 score over 98.70% in the classification of anomalous and normal requests. Furthermore, web attack detection time of our model is significantly lower (i.e., 0.4 ms) than the other approaches presented in the literature.
  • Küçük Resim
    Öğe
    Temporal transaction scraping assisted point of compromise detection with autoencoder based feature engineering
    (2021) Ogme, Fuat; Yavuz, Ali Gökhan; Güvensan, M. Amaç; Karsligil, Mine Elif
    Credit card fraudsters exploit various methods to capture card information. One of the common methods is to duplicate the credit cards by skimming. In this study, we introduce a new point of compromise detection method in order to trace and identify merchants where the skimming operation took place and card information has been captured by criminals. The proposed method first extracts discriminative features by using principle component analysis(PCA) and Autoencoder extractors and then it clusters similar fraudulent transactions with K-Means algorithm, afterwards it highlights possible merchants that are involved in this scheme by finding matching merchants in the produced clusters with a retrospective analysis of all transactions. Our experiments showed that the proposed method could achieve promising results with zeroknowledge on the existing skimming points. The application of our proposed method on real-life card transactions enabled us to pinpoint 7 out of 9 point of compromise previously identified by the reporting bank.
  • Küçük Resim
    Öğe
    Web application firewall based on anomaly detection using deep learning
    (2022) Toprak, Sezer; Yavuz, Ali Gökhan
    Anomali tespiti, farklı sektörlerde ve uygulama alanlarında araştırılmaya devam etmektedir. Anomali tespitindeki temel zorluk, benzersiz özelliklere ve yeni değerlere sahip bir girdi ile karşılaşılması durumunda normallerden aykırı değerleri belirlemektir. Araştırmalar, bu görevi yerine getirmek için Makine Öğrenmesi ve Derin Öğrenme tekniklerini kullanmaya odaklanmaktadır. Internet dünyasında, bir web sitesi isteğinin kötü niyetli veya sadece normal bir istek olup olmadığını belirlemek istediğimizde yine benzer bir sınıflandırma problemiyle karşı karşıya kalmaktayız. Web Uygulama Güvenlik Duvarı (WAF) sistemleri kötü niyetli faaliyetlere ve isteklere karşı, kural tabanlı ve son yıllarda kullanılan anomali tabanlı çözüm kullanarak koruma sağlar. Bu tür çözümler bir noktaya kadar güvenlik sağlar ve kullanılan teknikler, arka uç sistemlerini savunmasız bırakan hatalı sonuçlar üretmektedirler. Bu çalışmanın odak noktası, karakter sıralaması tabanlı bir LSTM (tekli ve yığılmış olmak üzere) yapısı kullanılarak bir WAF sistemi oluşturmak ve derin öğrenme modelinin optimum sonuç üretmesi için hiper parametrelerin hangi değerleri alması gerektiğini ortaya koymaktır. Semi-supervised öğrenme yaklaşımı için PayloadAllTheThings verisetinde yer alan gerçek saldırı verilerinin yanı sıra HTTP CSIC 2010 verisetinde yer alan ve normal olarak etiketlenen veriler hem modelin öğrenmesi sırasında hem de test edilmesi adımında kullanılmıştır. Önerilen tekniğin başarı oranının analizini için F1 skor değeri baz alınmıştır. Yapılan analizler ve deneyler sonucunda elde edilen derin öğrenme modelinin F1 başarı oranının yüksek olduğu ve saldırıları tespit etme ve sınıflandırma noktasında da başarı elde edildiği gösterilmiştir.