Sybil Attack Detection in IoT Device Using Machine Learning Algorithms

Abstract:

Sybil attacks are one of the most critical security threats to IoT networks, in which attackers use various fake identities to compromise or influence IoT systems. Traditional security methods such as encryption and authentication are unable to address Sybil attacks in the heterogeneous IoT environment. However, ML methods have been proposed for detecting Sybil attacks by analysing IoT data for patterns. This paper proposes a framework for detecting Sybil attacks in IoT networks using supervised ML methods such as Decision Trees, Neural Networks, and Gradient Boosting. Ensemble methods are also used for improved results. The framework uses datasets such as UNSW NB15 and KDD99 for testing. It uses robust pre processing methods, feature engineering, and k fold cross validation for testing. Results show high accuracy in detecting Sybil attacks. Ensemble methods achieve 99.9% accuracy. A prototype is also developed using Python and Flask for testing. Results show that the framework is scalable and adaptable to various IoT environments. Future enhancements to this framework will focus on developing real time Sybil attack detection and addressing emerging security issues in IoT.

Keywords: IoT security, Sybil attack, machine learning, network reliability, anomaly detection