Deep Learning Approach for IoT Vulnerability Severity Prediction

Le résumé fourni par la source

The proliferation of Internet of Things (IoT) devices has introduced a vast and complex attack surface, making efficient management and prioritization of vulnerabilities a critical cybersecurity challenge. Fast and accurate assessment of IoT vulnerability severity level is essential for effective prioritization and management of security risks. However, manual evaluation remains time-consuming, error-prone, and increasingly unsustainable given the rapid growth of IoT-specific vulnerabilities. This paper presents a novel approach that leverages a transfer learning model to automatically predict the severity of IoT vulnerabilities based on their textual descriptions, by leveraging the rich semantic information in vulnerability reports. This solution allows us to address the common issue of data scarcity in specialized domains like IoT security. Experimental results demonstrate that the proposed model significantly outperforms traditional machine learning techniques, achieving high accuracy of 91.5 and F1-scores of 0.90 in severity classification. This enables security teams to rapidly assess and prioritize vulnerabilities, thereby enhancing proactive defense measures and optimizing resource allocation in complex IoT environments.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.