Atleast 15% of users on social networking site Twitter are malicious users, said a study by a group of students of PEC University of Technology, Chandigarh.
The students worked on the accounts of nearly 50,000 users who had made their profiles public on Twitter. The accounts which were studied, had been created between the year 2006 and 2013 from across the world.
The accounts had been randomly chosen and all the data from these public accounts had consisted of people speaking English, Spanish, Arabic among others, said Shaleen Deep, one of the students who had worked on the project.
Nearly 10-lakh tweets by these users were analysed in the study. The malicious users are those who impersonate, spread viruses or spam the social networking site.
The study used only those accounts which chose to make their data accessible to avoid any privacy concerns.
The malicious users on the micro-blogging site was identified by analysing tweets by the users and checking how many people the user was following and how many followers the user had on Twitter.
To find out the malicious users, 5,085 suspicious users were identified among 50,000 users by scanning data of their tweets and profile patterns.In those suspicious users, 786 were found to be malicious.
"Since the time we began working on our model, 357 out of these 786 users accounts that had been identified as malicious using our model, were suspended by the Twitter in their internal exercise to delete spammers. It proved beyond doubt that they were malicious users," Shaleen Deep told Hindustan Times.
The study was carried out by Arjun D Sharma, Ansuya Ahluwalia and Shaleen Deep, all final year students of computer science department of PEC. Recently, the paper has been published in an international journal namely Springer LNCS.
"Twitter has shown considerable interest in this study. The research group at PEC is working on possibility of application of the model on a much larger data set provided by Twitter themselves," said an official of PEC.
The study also said there was a considerable number of shy and low profile users, who are genuine on Twitter and most of them used words like "my first tweet", "joined Twitter today", "feeling good and Hi Twitter".
The study says that malicious users are expected to maintain a very limited public presence on social networking. Their status and tweet count is also expected to be very low, especially over a period of time.
"The model can be helpful in detecting the malicious users on other social networking sites as well as they have same pattern of friends, followers and statuses," said the study in its conclusion.
How they arrived on this figure:
As many as 5,085 users were selected out of total 50,000 users whose activities were found to be suspicious after analysing their tweets and profiles. All the users were put on a scale of one to 10 on user score (how many followers, how many followings) and tweet score (how often the person tweets). Most of those found malicious users weighed between 0-1 on the scale.