Enhanced Dwarf Mongoose optimization algorithm with deep learning-based attack detection for drones

Abstract

Security in smart cities is a challenging issue in urban environments as they depend upon interconnected technologies and data for effective services. To address security challenges, smart cities implement robust cybersecurity measures, including network monitoring, encryption, and intrusion detection systems. Detecting and mitigating possible security risks in drone network B5G is a crucial aspect of ensuring reliable and safe drone operation. It is necessary to establish sophisticated and robust attack detection techniques to defend against security threats as the use of drones becomes increasingly widespread and their applications diversify. This is due to the lack of privacy and security consideration in the drone’s system, including an inadequate computation capability and unsecured wireless channels to perform advanced cryptographic algorithms. Intrusion detection systems (IDS) and anomaly detection systems can identify suspicious activities and monitor network traffic, such as anomalous communication patterns or unauthorized access attempts. Therefore, the study presents an enhanced dwarf mongoose optimization algorithm with deep learning-based attack detection (EDMOA-DLAD) in Networks B5G for the purpose of Drones technique. The presented EDMOA-DLAD technique aims to recognize the attacks and classifies them on the drone network B5G. Primarily, the EDMOA-DLAD technique designs a feature selection (FS) approach using EDMOA. To detect attacks, the EDMOA-DLAD technique uses a deep variational autoencoder (DVAE) classifier. Finally, the EDMOA-DLAD technique applies the beetle antenna search (BAS) technique for the optimum hyperparameter part of DVAE model. The outcome of EDMOA-DLAD approach can be verified on benchmark datasets. A wide range of simulations inferred that the EDMOA-DLAD method obtains enhanced performance of 99.79% over other classification techniques.