Large-scale IoT attack detection scheme based on LightGBM and feature selection using an improved salp swarm...

Area of Science:

• Cybersecurity
• Network Security
• Artificial Intelligence

Background:

• The rapid growth of Internet of Things (IoT) devices increases vulnerability to cyberattacks.
• Existing IoT attack detection methods struggle with accuracy, high dimensionality, and model efficiency.
• Metaheuristic algorithms offer a promising approach for effective feature selection in network data.

Purpose of the Study:

• To develop an efficient and accurate cyber threat detection scheme for the Internet of Things (IoT).
• To introduce an enhanced metaheuristic algorithm, GQBWSSA, for optimized feature selection.
• To create a lightweight ensemble learning model using LightGBM for improved IoT attack detection.

Main Methods:

• An enhanced metaheuristic algorithm, GQBWSSA (an improved Salp Swarm Algorithm), was developed.
• A threshold voting-based feature selection framework was designed using GQBWSSA to reduce data dimensionality.
• The selected features were integrated with the LightGBM algorithm to create an ensemble learning model for attack detection.

Main Results:

• GQBWSSA demonstrated superior performance in feature selection compared to existing metaheuristic algorithms.
• The proposed ensemble learning scheme achieved high accuracy, precision, and reduced training/detection times.
• On the CICIoT2023 dataset, the scheme reached 99.70% accuracy for binary classification and 99.41% for multi-class classification.

Conclusions:

• The GQBWSSA algorithm effectively addresses the curse of dimensionality and improves feature selection for IoT security.
• The proposed LightGBM-based ensemble learning model offers a lightweight and highly efficient solution for IoT cyber threat detection.
• The study validates the effectiveness of the proposed approach on both NSLKDD and large-scale CICIoT2023 datasets.