Department of Computer Engineering
CS 590 SEMINAR
Behaviour Monitoring: Detecting Compromised Twitter Accounts Based on Behaviour Changes
Hüseyin Celal Öner
Computer Engineering Department
Twitter is a fast-growing microblogging service that gives the power to share ideas and information instantly without barriers. It is an undeniable fact that like many popular internet services, Twitter is becoming a popular venue for the cyber-criminals each day. Malicious activities conducted by the cyber-criminals pose a major threat to its business and user experience. Threats against Twitter can be divided into two groups: fake accounts and compromised accounts. Fake accounts are mostly automated accounts which are primarily created to phish and spam the legitimate users. On the one hand, detection of fake accounts has been a subject of extensive studies in recent years, but on the other, detection of compromised account was not taken into consideration sufficiently.
In this work, we present a solution to the problem of detecting compromised accounts at Twitter. In order to address this problem, our system applies anomaly detection techniques. Firstly, by using historical user data the system learns the typical behaviour of individual users. Then, the system monitors for a suspicious deviation in user behaviour. Our goal is to successfully distinguish malicious and legitimate behaviour changes.
DATE: 10 April, 2017, Monday @ 16:55