One elementary issue in today’s on-line Social Networks (OSNs) is to relinquish users the power to manage the messages announce on their own non-public area to avoid that unwanted content is displayed. Up to now, OSNs offer very little support to the current demand. To fill the gap, during this paper, we have a tendency to propose a system permitting OSN users to ownan immediate management on the messages announce on their walls. This is often achieved through a versatile rule-based system and a Machine Learning-based soft classifier mechanically labeling messages in support of content-based filtering.