N-Gram Classifier System to Filter Spam Messages from OSN User Wall

Abstract

Online Social Network (OSN) allows users to create, comment, post, and read articles of their own interest within virtual communities. They may allow forming mini-networks within the bigger, more diverse social network service. But still, improper access management of the shared contents on the network may give rise to security and privacy problems like spam messages being generated on someone’s public or private wall through people like friends, unknown persons, and friends of friends. This may also reduce the interest of Internet data surfing and may cause damage to less secure data. To avoid this, there was a need of a system that could remove such unwanted contents, particularly the messages from OSN. Here in this paper, for secure message delivery I have presented a classifier system based on N-gram generated profile. This system consists of ML technique using soft classifier, that is, N-gram which will automatically label the received messages from users in support of content-based filtering. Effectiveness of N-grams is studied in this paper for the purpose of measuring the similarity between test documents and trained classified documents. As an enhancement, N-gram method can also be used to detect and prevent leakage of very sensitive data by using N-grams frequency for document classification.