Work

Integrating convolutional stacked long short-term memory with bio-inspired algorithm for secure unmanned aerial vehicle communication in SDN environment

Year: 2025

Type: article

Abstract: Unmanned aerial vehicles (UAVs), such as drones, are used in military operations on the battlefield to transmit information and assist strategic communication. Software-defined networking (SDN) presents an efficient approach to managing distributed wireless networks. Intrusion detection systems (IDS) are actively integrated into UAV systems using a novel SDN model designed for emergency scenarios. In these applications, preserving network security is an essential component to ensure reliable and efficient communication. UAVs equipped with IDS may play a crucial role in protecting and monitoring the communication structure in these applications. In this manuscript, a secure UAV-based real-time intrusion detection framework, enabled by Deep Learning and Bio-Inspired Optimisation (UAVID-EDLBIO), is proposed for the SDN environment. This method presents a novel network infrastructure and an original system that incorporates an advanced honeypot UAV to successfully identify and alleviate intrusion detection in SDN networks. Initially, the UAVID-EDLBIO approach employs Z-score normalization during data pre-processing to clean and transform raw data into a suitable format for analysis. For the feature selection process, the catch optimization fishing algorithm (CFOA) technique is employed to choose the most relevant features from the input data. Furthermore, the convolutional neural network and stacked long short-term memory (CNN-SLSTM) technique are implemented for the classification process. To further optimize the performance of the CNN-SLSTM classification model, the Marine Predators Algorithm (MPA) technique is employed for hyperparameter tuning, ensuring the selection of optimal parameters to enhance accuracy. To demonstrate the improved solution of the UAVID-EDLBIO technique, a comprehensive experimental analysis is conducted using the InSDN dataset. The comparison analysis of the UAVID-EDLBIO technique revealed a superior accuracy value of 99.88% compared to existing models.

Source: Results in Engineering

DOI: 10.1016/j.rineng.2025.106838