Web Based Anomaly Detection using Zero-Shot Learning with CNN

In recent years, cyberattacks have become a persistent threat, especially for websites. Therefore, web application security has significant issue in all industries under the evolution of intelligent devices and services. Dealing with imbalanced data is biggest obstacle to providing applications because there are less harmful despite large number innocuous request data. This paper suggests novel Zero-Shot Learning method employing Convolutional Neural Network address unbalanced high false positive rates (ZSL-CNN). approach uses only benign during training step, while it predicts unseen malicious requests. Three different datasets utilized comprehensive results. The first dataset containing internet banking logs provided by Yapı Kredi Teknoloji. Other open-source WAF HTTP CSIC 2010. After performing code embedding process, URI part obtained from these given as input ZSL-CNN model. outcomes then contrasted those using several methods, including Isolation Forest, Autoencoder, Denoising Autoencoder Dropout, One-Class SVM. being tested on mentioned above, experimental results demonstrate that true rate this model better than other reaching 99.29 %, respectively.