Detecting and Characterizing Arab Spammers Campaigns in Twitter
Reem M. Alharthi, Areej Alhothali, Kawthar Mostafa Moria · Procedia Computer Science · 2019
Social media platforms play a significant role in today’s society as billions of users use them to share or seek information on a regular basis. The content in social media has a significant impact that influences individuals’ opinions and decision, which can range from buying a small product to voting for a political campaign. This role has been fueling the interest in utilizing social media content for research and commercial purposes. The significant role of social media also attracted other individuals who intentionally abuse or misuse these platforms by producing and managing a tremendous amount of fake accounts to perform various malicious activities. Their activities are often center on sharing unwanted, misleading, or harmful content to either send unwanted ads, manipulate public opinion, or spread harmful malware. This act would eventually reduce the quality of social media content as a mean of information source used for various purposes. Therefore, this research proposes a machine-learning based model that aims to detect malicious users and groups on Twitter. We have, in particular, focus on malicious Arab accounts as it has not been sufficiently researched. The proposed model adopts a semi-supervised technique that labels Twitter accounts based on their behavior and profile information into spam or genuine account. To evaluate this technique, we collected a dataset through a Twitter API by targeting active Arab users. We manually labeled approximately 500 accounts as the ground truth, and then, we developed a classification model with a set of redefined features with the aim of identifying individual and groups of spam accounts. We have also evaluated the performance of the proposed model, and the results show that our model achieves 0.89 F measure: 0.89 and 0.91 Accuracy.