Trust represents an important issue for adopting cloud services. A trust management framework is essentially, about user rating. Hence, correctly addressing user feedback and filtering out malicious rating is a main step in providing reliable services. In order to process their feedback and calculate a reliable trust degree. Thus, new opportunities can be offered for the establishment of a trust relationship among involved entities. We propose a technique to process user rating by statistical methods. Then, we proceed to classify the users into different groups to detect malicious users. The users are grouped according to their rating by a k-means clustering technique, and the evaluation will show that the proposed solution gives better results than the traditional filtering solution.