Nesnelerin İnternetinde Sahte Kimlik Saldırılarının Makine Öğrenme Yöntemleri ile Tespiti

Abstract

Internet of Things (IoT) devices are increasing their usage rates with advances in the wireless sensor networks. All IoT devices are connected to themselves with a heterogeneous network. Thus, they are also rather vulnerable to external attacks. Many routing protocol attacks have been described until now and continue to expand and diversify. Therefore, the recommended detection and prevention methods should be updated and improved according to today’s condition. Sybil attack is a kind of the Routing Protocol for Low-Power and Lossy Network (RPL) attacks in IoT. The attack detection based on the signal strength of the nodes in Sybil attacks are one of the most commonly used and recommended approaches. In particular, classical methods that used to detect and prevent attack may not be appropriate for attack detection. The most critical problems in resource constrained IoT systems are energy consumption and heavy computational cost. In this study, packet distribution rates and machine learning approaches such as Naïve Bayes, Random Forest and Logistic Regression have been proposed for the prediction of Sybil attacks on RPL protocol in IoT networks. The Sybil attacks have been detected with 99.51% accuracy rate and this result is higher than classical methods for Sybil attack detection.