Abstract: Cloud computing and Machine learning are the widely used technologies that are grabbing the attention of every user of technology. The cloud users are under constant threat of losing crucial data to the unauthorized users. With the advent of machine learning, learning based approaches like supervised and unsupervised are becoming more popular in building security framework for clouds. Machine learning algorithms used for the purpose of security are trained with the datasets and they are tested with different datasets. For this part of research work we have used comprehensive datasets to provide robustness to the models. Confidentiality and Integrity are the two major concerns of data security. In this paper we have compared five supervised machine learning algorithms to classify the data before providing data confidentiality. Logistic Regression, Decision Tree, Random Forests, Extra Trees, KNN are the five ML algorithms that are applied on the data sets to classify the data. Data is classified as critical or non critical in order to reduce the computational time for data encryption. Not all the data is encrypted but only the critical data which means to be secured should be converted to cipher. To increase security of the data and reduce computational complexity of the security framework, our work focuses on using ML algorithms for data classification.

Logistic Regression, Decision Tree, Random Forests, Extra Trees, KNN, AES