The integration of the Web of Things with cloud computing platforms has significantly improved data sharing, analytics, and automation across smart environments. However, this interconnection also exposes devices and cloud infrastructures to severe security threats, including Trojans, ransomware, worms, and advanced malware. Ensuring the confidentiality, integrity, and availability of data in such ecosystems requires intelligent and adaptive threatware detection mechanisms. Problem Statement Conventional detection systems based on static rules, signatures, or isolated behavioral analysis fail to recognize evolving threat patterns and zero‐day attacks. These methods also suffer from high computational cost and limited accuracy when applied to dynamic Web of Things cloud environments. Need for Research A robust, intelligent, and optimized detection framework is essential to identify complex and unknown threatware efficiently while minimizing computational overhead and false alarms. Proposed Work and Objective This research proposes the metaheuristic optimization algorithm using enhanced deep learning for threatware analysis model. The primary objective is to automatically detect and classify threatware in Web of Things cloud architectures through optimized feature selection and adaptive deep learning. Novelty The uniqueness of the proposed model lies in combining population‐based metaheuristic optimization for hyperparameter tuning with a deep sequence‐to‐sequence recurrent neural network for intelligent classification. This integration improves both detection speed and classification precision. Method The model incorporates metaheuristic‐driven attribute selection and enhanced deep neural learning to analyze data sequences and identify threatware behaviors accurately. Dataset The threatware dataset (curated benchmark dataset) containing 4200 labeled samples across nine categories was used for performance evaluation, ensuring diversity and real‐world representativeness. Results The proposed model achieved an accuracy of 98.72% on the training set and 99.09% on the testing set. It also recorded the lowest computation time of 7.12 s compared with existing approaches such as Artificial Algae Optimization with Deep Belief Network, convolutional neural network–long short‐term memory, and bidirectional recurrent neural network. Conclusion and Future Work The findings confirm that the proposed model enhances detection accuracy, reduces computational time, and strengthens threatware resilience in Web of Things cloud environments. Future work will explore larger datasets, real‐time adaptive learning, and hybrid metaheuristic optimization to further improve system scalability and robustness.
Rawat et al. (Thu,) studied this question.