Abstract

With the growing popularity of social media, social network (you tube) remains the largest as well as the most popular video sharing site. However, terrorists groups have made YouTube as a focal point for targeting innocent and vulnerable people. They propagate their ideologies to mainstream audience who otherwise would not visit their website. Hence, there is a need to detect such videos to prevent online radicalization among the users. The paper proposes a metadata and audio based classification method for detecting such videos which promote hate and violence by mining the user generated metadata such as title, description which the uploader of the video adds along with finding patterns to classify an audio into violence class such as gunshots and screams.