Cyber vulnerabilities detection system in logistics-based IoT data exchange

Abstract

Modern-day digitalization has a profound impact on business and society, revolutionizing logistics. Supply chain digitalization improves transparency, speed, and cost-effectiveness, increasing tech adoption—transportation benefits from IoT-driven shipment tracking and web data storage. However, cyber threats target IoT data by exploiting cyber vulnerabilities. Although ML/DL approaches have showed potential in finding IoT vulnerabilities, the difficulty of selecting appropriate features remains. Existing research has produced surprising outcomes, and deep neural networks have been utilised to extract characteristics without taking sequence information into account. To address this, the paper presents a unique approach for accurate IoT vulnerability identification that combines deep learning and better feature selection. On the BoT-IoT dataset, the LSTM + CNN model achieved 95.73 % accuracy. This approach has the ability to successfully anticipate IoT based vulnerabilities by leveraging benchmark data, selecting relevant features, and enhancing overall system performance.

Keywords

• Hybrid deep learning
• Feature selection
• IoT-based vulnerabilities
• Logistics