TaWSA-WRN: a Taylor wave search optimized WideResNet framework for intrusion detection with response-aware mitigation in cloud computing.

Targeting cloud computing environments has become increasingly attractive to sophisticated cyberattackers due to their open, scalable, and distributed nature. Intrusion Detection Systems (IDSs) analyse traffic patterns to detect these attacks; deep learning has become a common approach for building such systems, but many suffer from overfitting, high false-positive rates, and/or unstable training behaviour. To overcome these drawbacks, this paper introduces a Taylor Wave Search Algorithm-optimised Wide Residual Network (TaWSA_WRN) framework for intrusion detection and mitigation in cloud environments, which protects against incoming attacks. The network traffic data accessed from benchmark datasets, such as NSL-KDD and CICIDS2017, are subsequently preprocessed by restoring missing values and applying Min-Max normalisation. Finally, the TaWSA_WRN model performs feature selection and intrusion classification, using the Taylor Wave Search Algorithm to enhance parameter optimisation and improve learning stability by optimising a hyperparameter. This model-agnostic interpretability approach, using SHAP, also offers valuable insights into domain-relevant traffic features. The empirical results show that the proposed approach can achieve maximum TNR, accuracy, and TPR of 96.857%, 97.190%, and 97.589%, respectively. We introduce the mitigation component as a response-guided strategy based on detections rather than a single network-level defence. The proposed framework enables greater reliability, interpretability, and robustness, thereby facilitating detection in a secure cloud computing environment. Code implementation is available at https://github.com/vedasahithi/wideResNet-for-intrusion-detection -.