An Efficient Hybrid Feature Select Technique towards Prediction of Suspicious URLs in IoT Environment

Abstract

the evolving landscape of the Internet of Things (IoT), securing network-connected devices from cyber threats has become a critical concern. Among these threats, malicious URLs pose a significant risk by facilitating phishing, data theft, and malware attacks. This paper proposes an efficient hybrid feature selection technique aimed at enhancing the prediction of suspicious URLs within an IoT environment. The hybrid approach combines filter and wrapper-based methods to extract the most relevant features from large and complex URL datasets. By applying machine learning classifiers such as Random Forest and Support Vector Machines (SVM), the system demonstrates improved accuracy, reduced false positive rates, and faster detection times. The proposed technique optimizes feature space, reduces computational cost, and increases prediction reliability, making it highly suitable for real-time threat detection in resource-constrained IoT devices.