COURSTEYNov 07 2016 : The Times of India (Chandigarh)
PEC scholar's site to check true Twitter followers
Vishakha Chaman
Chandigarh


Ever wondered who comprises your expanding base of followers on Twitter? A research scholar at PEC University of Technology , Chandigarh, is all set to launch a website where the Twitter users will be able to check whet her their follo wers are genu ine or spam mers.
Monika Singh, a research scholar at PEC, along with her supervisors Divya Bansal and Sanjeev Sofat, has prepared a framework to detect five kinds of spammers--pornographic spammers, compromised profiles, Twitter followers' user merchants, fake profiles and sole spammers.

In a paper titled “behavioural analysis and classification of spammers distributing pornographic content in social media“ published in Springer, Monika has analysed 74,000 tweets containing `adult content' from 18,000 Twitter users in the first part of the study. “I have used the tweet content and demography of the users to detect the category of spammers they fall in. After analyzing the data, I used a machine learning algorithm to check the accuracy of the findings; and found it to be 84.4% accurate,“ said Monika. She has used six phrases related to pornography to detect the pornographic content in tweets.

This study identified around 2,000 Uniform Resource Locators (URLs) used by pornographic spammers to disseminate adult content on Twitter. The characteristics used to conduct the behavioural analysis of these Twitter users fall under the stated spam policy of Twitter.

The second part of the study, which is under review, deals with the analysis of the Twitter Followers Mar ket Merchants -who sell followers on Twitter.

Monika claims that she is the first one to have analysed this category of users who are actually disseminating spams on Twitter. “I have analysed 995 users and 16,000 tweets.“

Current Twitter rules recommend that users not use any pornographic content in either profile or header image.The third part includes developing `real-time malicious user detector', the application of the study she has conducted by collecting data of 1 lakh Twitter users and 20 million tweets. However, validation has been done on 20,000 tweets and the overall accuracy of this third part of the work is 92.1% using machine algorithm