Ahmetoglu, H.Das, R.2025-02-152025-02-1520199781728129327https://doi.org/10.1109/IDAP.2019.8875872https://hdl.handle.net/20.500.12514/6231One of the solutions proposed to ensure information security is intrusion detection systems. Improving the performance of these systems has been among the most important objectives of information technologies. In this study, a detailed analysis of the explicitly presented CICIDS2017 data set was performed. The data set was rearranged by collecting different types of attacks under the same heading for binary classification. For multiple classifications, all files it contains are combined. Using the new version of the data set, a sample model has been developed with the Full Linked Artificial Neural Network, which is one of the machine learning techniques. This model is encoded with TensorFlow-Keras libraries and classified using network traffic properties. The success of the dual classification results and the multiple classification successes were compared. Multiple classification can include the type of attack. On the other hand, in case of dual classification, the attack is present and no attack status is examined. The success rate of binary classification is expected to reduce false alarm conditions in intrusion detection systems. © 2019 IEEE.tr10.1109/IDAP.2019.8875872info:eu-repo/semantics/closedAccessArtificial Neural NetworksDeep LearningIntrusion Detection SystemMachine LearningConference ObjectN/AN/A2-s2.0-850748868783