Hybrid ANN–DNN–LSTM Deep Learning Architecture for Botnet Detection in Smart Grid Cyber-Physical Systems

Abstract

Smart power grids are vital cyber-physical infrastructures that are more vulnerable to sophisticated cyberthreats, especially botnet-driven assaults, due to their growing reliance on IoT-enabled communication and control technology. These assaults have the potential to spread quickly across both physical and cyber levels, resulting in serious stability threats, operational interruptions, and cascade failures. Therefore, rapid and accurate intrusion detection is crucial since false negatives in power systems might have permanent physical repercussions. However, current intrusion detection methods based on standalone deep learning models frequently have poor handling of the temporal and nonlinear dependencies present in smart grid data, poor generalization, and high false-negative rates. To address these limitations, this study presents a hybrid ensemble deep learning framework that integrates Artificial Neural Networks (ANN), Deep Neural Networks (DNN), and Long Short-Term Memory (LSTM) models for enhanced botnet attack detection in smart grid cyber-physical environments. Through an ensemble-based decision technique, the framework makes use of the complementary characteristics of the component models, temporal learning, hierarchical feature extraction, and nonlinear pattern modeling. A benchmark smart grid dataset from the Kaggle Machine Learning Repository is used in a three-phase technique that includes data pretreatment, model integration, and performance evaluation. Experimental results demonstrate that the hybrid ensemble significantly outperforms individual models upon evaluation, with a marked reduction in false negatives. These findings confirm the effectiveness of ensemble hybridization in improving detection reliability and resilience, positioning the proposed framework as a high-fidelity supervisory mechanism for securing smart grid cyber-physical systems against evolving botnet threats.