Gökdemir, AliÇalhan, Ali2023-07-262023-07-2620221300-18841304-4915https://doi.org/10.17341/gazimmfd.962375https://search.trdizin.gov.tr/yayin/detay/508765https://hdl.handle.net/20.500.12684/13378Graphical/Tabular Abstract Classical machine learning and deep learning were compared in detecting attacks on IoT environments. Due to its success in anomaly detection in the literature, Support Vector Machines (SVM) and Naive Bayes (NB) algorithms from classical machine learning algorithms were preferred. As a deep learning algorithm, the Long Short-Term Memory (LSTM) algorithm, which is mostly used in fields such as natural language processing and text processing, and which has very few studies in anomaly detection, has been chosen. With the LSTM algorithm, higher values were obtained in accuracy and f1 scores. Figure A. Proposed system model for anomaly detection in IoT environments with LSTM-SVM-NB algorithms Purpose: As the use of Internet of Things (IoT) systems has become widespread, cyber-attacks against these systems have also increased. Cyber-attacks occurring in IoT environments can include different types of attacks, such as the inability of their devices to serve, corruption, data capture, modification, or deletion. In this study, it is tried to predict duplication, interception, and modification attacks in Message Queuing Telemetry Transport (MQTT) messages using an IoT dataset with artificial intelligence techniques. Theory and Methods: In this study, compared to the performance metrics of SVM and NB, which are machine learning algorithms, and LSTM, which is a deep learning algorithm. Results: Experimental results show that the LSTM algorithm can be used in anomaly detection in the cyber security area, apart from natural language processing and text processing, which are the areas widely used in the literature. Besides, it was concluded that the LSTM algorithm achieved higher accuracy than the classical machine learning algorithms. Conclusion: In this paper, a comparison of deep learning and machine learning algorithms for anomaly detection in IoT environments is made. The results show that the LSTM algorithm, gives more effective results in anomaly detection than classical machine learning algorithms, but has some disadvantages in terms of working time.en10.17341/gazimmfd.962375info:eu-repo/semantics/openAccessIot; Machine Learning; Deep Learning; Anomaly; Iot SecurityAttacks; SchemeArticle374194519565087652-s2.0-85128770373WOS:000767316300017Q2Q4