Abstract
The usage of social media is increasing rapidly in the Arab region. One of the popular social networking sites for sharing news and spreading propaganda is Twitter. Spammers use these sites to disseminate adult content and false political news in Arabic tweets. Within the Arab region, distributing adult materials is illegal and some governments attempted to block malicious URLs. In this paper, we study both user and content attributes to differentiate between legitimate and illegitimate users. Then, we use those attributes with machine learning algorithms to detect spam on Twitter. We use Naive Bayes (NB) and Support Vector Machine (SVM) classification methods to find malicious contents in the tweets. Our results show that NB produces more accurate outcomes for detecting spam in Gulf Dialectical Arabic tweets.