This paper proposes a new method of features extraction for handwritten, printed and isolated numeral recognition. It is essential today for a company to store its data in an encrypted way when it uses Cloud Computing. However, the manipulation of this encrypted data remains complex, and it is very difficult in this case to be able to share the encrypted data between different users. One of the solutions for sharing encrypted data is to use PRE (Proxy Reencryption) which allows both the re-encryption of the data, but also the delegation of this operation by a third party via the use of a specific key. In this article, we propose a solution for sharing encrypted files between users that uses a classic storage system in the Cloud and PRE (re-encryption PRoxy). We present an improvement of an existing PRE algorithm by applying it to elliptical curves in order to improve its performance. Finally, we implement this architecture in the form of a cloud service called PREaaS (PRE as a Service) which allows this mechanism to be used on demand with an API.