Identification of Malicious Bots in Twitter using Wavelets

Abstract

A social network is a platform for users to express their ideas and opinions towards the world. The social network is activated across various computer networks. By using Online social network (OSN) a user can share digital photos/videos/texts to their friends. Users in OSN can perform fast communication between each users. The proposed method classify twitter accounts as human legitimate bot or malicious bot using tweets. Most of the previous works focus on classify twitter accounts to two: bot or human. But bots may be harmful or harmless. Harmless bots (legitimate bot) doesn’t leads to any malicious activities on users; sometimes it cause traffic problems. Actual goal of legitimate bot is similar to the goal of social media that is, spreading news. So there is a need to classify bot account to two malicious bot and legitimate bot. Here we use wavelet properties for the tweet document and find features like, corpus size, lexicon size, URL count and wavelet domain score from the document. All the selected features are applied to supervised machine learning algorithm and classify the tweet as human, legitimate bot or malicious bot. Best algorithm is selected by comparing their accuracy.