The authors of this paper introduce a novel method of IoT security with the applied integrated framework based on a Tunicate Swarm Algorithm with stacked deep learning models. Because various security risks have begun appearing due to the IoT network, the suggested solution integrates the optimization feature of the Tunicate Swarm Algorithm and the powerful classification techniques of deep learning methods to form a combined framework that can detect and address the security weaknesses more effectively. In the given paper, we demonstrate that the performance has indeed increased, with the help of detailed experiments and measurements of the metrics on the basis of typical datasets provided by the internet of things security. This, therefore, highlights the possibility of such a hybrid solution to a strong and scalable security of the IoT network, which remains in the absence up to the present time, and thus a basis to the further research and practical applications in other IoT settings.