Secure private cloud using machine learning and cryptography

Abstract

There is a continuous increase in the use of the private cloud due to its many and important features. There are increase security challenges that target these systems. One of the most important requirements of users in cloud storage is to protect their cloud from attacks, and keep data secret. Modern technologies of machine learning provide the ability to analyze and classify the data. This paper proposes a model that serves as a server (a third party) between private cloud and users and it is based on two phases. The first of which is to protect the cloud from different types of attacks and detect normal and abnormal flow. Whilst the second one is to categorize the users’ data to be stored in the cloud and then encrypt it based on the degree of its importance using different types of encryption algorithms. The results show the accuracy respectively as Naive Bays 74%, SGD 82%, LR 85%, KNN 89%, and Random Forest 100%, Decision Tree 100%. For the second phase of classifying the data, the algorithms used are the decision tree, KNN, Random Forest, Naive Bays, LR, and SGD. The results have respectively 82%, 93%, 96%, and 97%, 98% and 98% respectively. Finally, the Random Forest and SGD algorithms for the first and second phases were used. The encryption algorithms that have been adopted are RC4, 3DES, and AES for encryption of classified data according to the importance of the second phase to be stored in the cloud in a secure form.