Lightweight Intrusion Detection in Cloud Computing with Squeezed Excitation Block and RUNge Kutta Optimizer

Cloud computing has transformed data storage and online services, but its increasing usage has raised concerns about intrusion threats. Effective intrusion detection is critical to ensuring the security of these environments. In this context, this paper proposes a low-cost and lightweight intrusion model for cloud computing. The proposed model used RUNge Kutta Optimizer (RUN) to extract the most relevant features from the incoming traffic. After this, the squeezed and excitation (SE) block is used to add channel-wise attention, improving feature recalibration and boosting model accuracy. The use of SE block makes the proposed model lightweight (933 trainable parameters (TP)) as compared to standard deep learning models like CNN (5989 TP), GRU (20005 TP), LSTM (39255 TP), RNN (10005 TP). The proposed model also outperforms the other deep learning model with an accuracy of 0.98046 and precision of 0.98056. As the proposed model is efficient, lightweight, and economical, it addresses the dynamic security needs of intelligent cloud systems.