Cyber security requires detecting and stopping wireless sensor network attacks. We introduce a fast super deep learning model for cloud-based WSN attack detection in this research. Our offline approach divides data into training and testing sets, preprocesses it, then reduces features using Singular Value Decomposition (SVD) and Principal Component Analysis. We then extract features using our ultrafast deep neural network and put them into a hyper deep-machine learning model. We evaluate our strategy using data mining metrics and preserve the model for further use. We also compare our model to other data mining and machine learning methods using numerous datasets to ensure accuracy and efficiency. We focus on a few key topics. First, we address generalization, complexity, and execution computation time. Second, many datasets increase model generalization. Third, our fast and accurate deep learning approach improves machine learning results. Our Oracle and Java cloud infrastructure lets us design an injection tool to produce and transmit packets to the intrusion detection model. Model predictions stored on an Oracle database offer complete attack information. Convolutional Neural Networks (CNN) and machine learning algorithms like Complement Naive Bayes (CNB) and Stochastic Gradient Descent are among our methods. We also study combining PCA and SVD with machine learning to improve results. Finally, deep feature selection improves the previously reported methods. Overall, our proposed technique for cloud-based WSN attack detection has high generalization, efficiency, and accuracy while addressing major cyber security challenges.