DETECTION OF ATTACKS ON WEB APPLICATIONS BASED ON A COMBINATION OF CONVOLUTIONAL NEURAL NETWORKS AND MULTILAYER PERCEPTRONS
I. V. Kotenko, Pavel Sobolev · Informatization and communication · 2025
Bi.Zone company specialists analyzed vulnerabilities found in Russian and foreign web applications at the end of 2024. According to these data, about a thousand new web vulnerabilities are discovered in the world every month. Therefore, ensuring the security of web applications remains a critical cyber security task. This paper proposes a method for detecting attacks on web applications using a combined approach based on deep learning, including convolutional neural networks (CNN), recurrent neural networks (LSTM), multilayer perceptrons (MLP) and their combinations in order to increase the accuracy of detecting attacks on web applications. The main focus is on the CNN and MLP hybrid model, which combines the advantages of spatial analysis and powerful classification capabilities. The conducted testing showed the effectiveness of the approach on a number of datasets, including CSIC 2010 and various Kaggle datasets.