Deep learning based phishing web page detection
Files
Date
2022
Authors
Bastem, Tevfik Uğur
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
E-ticaret, sosyal medya ve dijital hizmetlerin kullanımının artmasıyla birlikte oltalama faaliyetlerinde muazzam bir artış yaşanmaktadır. Bu çalışmada, oltalama faaliyetlerinin gözlemlenmesinden yola çıkarak, transfer öğrenme yöntemleri ile sahte web sitelerinin tespitini yapacak bir çalışma gerçekleştirilmiştir. Çalışma kapsamında, adobe, amazon, apple, microsoft gibi web sitelerinin gerçek ve sahte ekran görüntülerinden oluşsan toplam 2852 ekran görüntüsü içeren bir veri setinden yararlanılmıştır. AlexNet, VGG16, RESNET50 transfer öğrenme yöntemleri yanı sıra kendi geliştirdiğimiz çok girişli CNN modelini kullanarak, elde edilen sonuçlar analiz edilmiştir. Elde edilen bulguların gelecekte yapılabilecek diğer çalışmalara etkisi tartışılmıştır.
With the increase in the usage of e-commerce, social media and digital entertainment services, there is a tremendous increase in phishing activities. In this study, based on the observation of phishing activities, a study has been carried out to detect phishing websites with deep models and transfer learning. Within the scope of the study, a data set containing a total of 2852 screenshots, consisting of real and fake screenshots of websites such as adobe, amazon, apple and microsoft etc. was used. The results obtained by using transfer learning from AlexNet, VGG16 and RESNET50 models as well as the proposed multi- input CNN model were analyzed. Promising results are obtained from the experiments. The effects of the obtained findings on other future studies were discussed.
With the increase in the usage of e-commerce, social media and digital entertainment services, there is a tremendous increase in phishing activities. In this study, based on the observation of phishing activities, a study has been carried out to detect phishing websites with deep models and transfer learning. Within the scope of the study, a data set containing a total of 2852 screenshots, consisting of real and fake screenshots of websites such as adobe, amazon, apple and microsoft etc. was used. The results obtained by using transfer learning from AlexNet, VGG16 and RESNET50 models as well as the proposed multi- input CNN model were analyzed. Promising results are obtained from the experiments. The effects of the obtained findings on other future studies were discussed.
Description
Keywords
AlexNet, VGG16, RESNET50, Transfer Learning, Multi-input CNN, Phishing, Transfer öğrenimi, Çok Girdili Sinir Ağı, Oltalama
Citation
Bastem, Tevfik Uğur (2022). Deep learning based phishing web page detection / Derin öğrenme tabanlı kimlik avı web sayfası tespiti. Yayımlanmış yüksek lisans tezi. Ankara: Çankaya Üniversitesi Fen Bilimleri Enstitüsü.