Enhanced Suspicious URL Detection in IoT Using an Optimized Hybrid Selection Technique

The rapid growth of the Internet of Things (IoT) has increased the threat of data breaches from malicious links. Identifying suspicious URLs before access is essential for protecting sensitive information. Machine learning methods are effective in detecting zero-day attacks, but their success relies on the quality and complexity of selected features. Earlier approaches primarily used lexical features for faster detection but failed to provide comprehensive website analysis. Enhancing IoT security requires combining both lexical and page content-based features. Researchers use various Feature Selection Techniques (FSTs) to extract meaningful features. However, high resource demands and complex datasets have led to the development of hybrid FSTs. The proposed hybrid FST integrates a filter-based method with a Genetic Algorithm (GA), enhancing the identification of malicious URLs and links. It leverages diverse feature sets and optimized boosting estimators to improve detection accuracy. The model achieves 99% accuracy while minimizing computational costs. This approach strengthens the security of IoT networks by addressing the limitations of previous methods. Efficient feature selection and boosting techniques ensure quick and  making it ideal for resource-limited IoT devices.