Cloud computing plays a major role in the business domain today as computing resources are delivered as a utility on demand to customers over the Internet. Cloud storage is one of the services provided in cloud computing which has been increasing in popularity. The main advantage of using cloud storage from the customers' point of view is that customers can reduce their expenditure in purchasing and maintaining storage infrastructure while only paying for the amount of storage requested, which can be scaled-up and down upon demand. With the growing data size of cloud computing, a reduction in data volumes could help providers reducing the costs of running large storage system and saving energy consumption. So data deduplication techniques have been brought to improve storage efficiency in cloud storages. With the dynamic nature of data in cloud storage, data usage in cloud changes overtime, some data chunks may be read frequently in period of time, but may not be used in another time period. Some datasets may be frequently accessed or updated by multiple users at the same time, while others may need the high level of redundancy for reliability requirement. Therefore, it is crucial to support this dynamic feature in cloud storage. However current approaches are mostly focused on static scheme, which limits their full applicability in dynamic characteristic of data in cloud storage. In this paper, we propose a dynamic deduplication scheme for cloud storage, which aiming to improve storage efficiency and maintaining redundancy for fault tolerance.