Secure Integration of Internet of Things with Cloud Computing Using Optimized Feature Selection and Convolutional Neural Network

This paper introduces a robust deep learning frame-work designed for the secure integration of Internet of Things (IoT) with cloud computing, aimed at enhancing cyber defense mechanisms. By leveraging an optimized feature selection approach and a Convolutional Neural Network (CNN), our model effectively filters attack traffic, addressing critical security challenges in IoT-cloud systems. We validate our model against the KDD Cup dataset, employing a two-phase process that initially reduces feature dimensions, followed by malicious traffic identification. Our results, demonstrated through accuracy, loss comparisons, and a confusion matrix, indicate a notable outperformance over traditional models such as Logistic Regression, Simple Neural Networks, and Support Vector Machines (SVM). The proposed model achieves a classification accuracy of 97%, with a confusion matrix analysis reinforcing its precision and recall strengths. The training phase shows our model's rapid convergence and stability, confirming its potential as a scalable solution for advanced cybersecurity applications.