Data is a valued asset especially when data moving to the cloud. Cloud computing model attracts different users because of its various benefits such as scalability, simplicity, expense reduction, and high resource resistance. Cloud computing is a wide-ranging architecture based on different models for providing different software services and hardware services. There may be different types of data and the degree of protection needed for all the data is also may vary. Data privacy and data security is the major area of research and development in the cloud computing. Privacy protection and data leakage became essential thing for so many organizations moving on to cloud. Several classification methods that are expressed various parameters based on different dimensions for providing the data security can be based on the level and are required protection. But, the latest approaches introduced by the cloud, related to distributed resources, multi-tenancy idea, computation outsourcing, high dynamism of the model, data warehousing and the non-transparent style of cloud increase the security and privacy concerns and makes building and maintaining trust among cloud service providers and consumers is a serious security issue. In this paper, we proposes a new approach to improve security of data in cloud computing. It proposes a classification model to classify data before being presented into a suitable encryption system based on the category. Because data in cloud has not the same sensitivity level, encrypting these different types of data with the same algorithms can lead to a lack of security or resources. By this scheme, we can achieve high security for our data, and can optimize the computation cost and resource consumption while ensuring data confidentiality.The efficiency of the proposed classification scheme is analysed with the sample dataset collected.