o:3009 Veštačke neuronske mreže za detekciju veb napada Artificial neural networks for web attack detection : doctoral dissertation sr This dissertation presents a comprehensive approach to web attackdetection using artificial neural networks. The collection of diversemalicious web traffic was carried out using honeypots, enabling thecreation of a robust dataset for training models to identify cyberthreats. The study addresses zero-day attack detection through variousmachine learning techniques capable of identifying previouslyunknown vulnerabilities in network traffic. To reduce catastrophicforgetting in dynamic attack environments, the use of multipleincremental learning strategies has been proposed, which enablecontinuous model adaptation with minimal loss of previouslyacquired knowledge. Тhe study introduces a population-based featureselection method, which improves classification efficiency byfocusing on the most relevant network features. The dissertationpresents a deep learning model for phishing email detection, based onthe architectures of recurrent and convolutional neural networks.Moreover, advanced feature weighting and embedding techniques areemployed to enhance phishing website detection. By integrating thesemethods, this dissertation provides a scalable and adaptive solutionfor real-time detection of web-based threats, offering significantadvancements in the fields of web security and machine learning. Biografija autora: str. [103].Bibliografija: str. [92-102]. Datum odbrane: 15.07.2025. Artificial intelligence 91552100 173068041 91552101 8783 2025-10-01T12:58:46.053Z 45 no 46 Nikola M., 1992 Stevanović 127609865 2025 63 mentor Branimir, 1967- Todorović 28413287 2025 63 član komisije Miroslav, 1964- Ćirić 13733223 2025 63 član komisije Marko, 1979- Miladinović 28396391 2025 63 član komisije Aleksandar B., 1989- Trokicić 79656969 2025 63 član komisije Dragan, 1967- Janković 3739495 2025 101 list 5906441 http://phaidrani.ni.ac.rs/o:3009 no yes 12 2025-10-01T12:58:46.320Z 70 veštačke neuronske mreže, rekurentne neuronske mreže,konvolucione neuronske mreže, odabir atributa, ponderisanjeatributa, inkrementalno učenje, sajber bezbednost, detekcija vebnapada, zamke, detekcija fišinga artificial neural networks, recurrent neural networks, convolutionalneural networks, feature selection, feature weighting, incrementallearning, cybersecurity, web attack detection, honeypots, phishingdetection 004.8(043.3) P 176 1738 18A07 18A0701 2025