Abstract: Now a dayís, people can communicate with each other by exchanging the multimedia information. It may include textual data, images, audio or video. Online social Network (OSN) provides different services to the user. In online social networking sites like facebook, twitter, etc., there is possibility of posting any kind of data on user wall. Such data may contain unwanted messages or images. Other user can view such data and can comment on such post. Such post may affect user social image. So security of such userís personal wall is an important issue. So, to avoid this, we propose a system which provides facility to user to control the unwanted messages and images posted on their wall and a system that allow users to restrict particular user on the basis of its social reputation, also user can extract labels from posted images and filter it, also take a decision whether to allow such content or not. This system also suggests malicious users from their behaviour and block frequency. Hence, this system, avoids nuisance of unwanted messages and images on the walls.

Keywords: online social networks, short text classifier, message filtering, labelled image, machine learning.