Background: With the entry into the world of cloud computing, which has become an essential in our lives, especially after the pandemic of the era of COVID-19, and with the increase of overlapping data over the Internet and networks with an increasing and great speed, the need to protect those data and applications, especially as the usage of cloud computing, increases. Objectives: Searching for the best solutions to provide the necessary protection against data attacks via cloud computing, so the need has become more urgent to access huge storage resources and applications with ease and process them anywhere with flexibility and more security. Methods: In this paper, a large group of images has been encoded using one of the encryption algorithms, then we used the Convolutional Neural Network (CNN) algorithm, which is a widely applied deep learning technique for image recognition. There is no doubt that deep learning has many models that contribute to increasing the speed and accuracy in the appearance of the results, including the ResNet50 model, where we took the model by training many encrypted images. Through this model, it was able to classify and identify the encrypted images without decoding them. Results: It was found that the classification of encoded images using the deep learning technique of the ResNet50 model has the potential to identify the encrypted image without decoding it. The proposed model achieved accuracy (99.75%), Recall (94.12%), Precision (94.23%) and F1-score (94.70%) on the test dataset, which indicates the feasibility of this approach in classifying encoded images.