Detecting Malicious Twitter Bots Using Machine Learning

Cybercrimes and phishing scams have increased multi-folds over the past few years. Now a days, hackers are coming up with new techniques to hack accounts and gain sensitive information about people and organisations. Social networking site like Twitter is one such tool. And due to its large audience hackers use such sites to reach large number of people. They use such sites to circulate malicious URLs, phishing mails etc. which serve as the entry point into the target system. And with the introduction of Twitter Bots, this work got even easier. Twitter bots can send tweets without any human intervention after a fixed regular interval of time. Also their frequency of tweets is much more than humans and therefore they are frequently used by hackers to spread malicious URLs. And due to large number of active members, these malicious URLs are reaching out to more people, therefore increasing the phishing scams and frauds. So this paper proposes a model which will use different algorithms of machine learning, first to detect twitter bots and then find out which of them is posting malicious URLs. In the proposed model, some features have been suggested which distinguishes a twitter bot account from a benign account. Based on those statistical features, model will be trained. The model will help us to filter out the malicious bots which are harmful for legitimate users.