Statistical Twitter spam detection demystified: performance, stability and scalability

With the trend that the Internet is becoming more accessible and our devices being more mobile, people are spending an increasing amount of time on social networks. However, due to the popularity of online social networks, cyber criminals are spamming on these platforms for potential victims. The spams lure users to external phishing sites or malware downloads, which has become a huge issue for online safety and undermined user experience. Nevertheless, the current solutions fail to detect Twitter spams precisely and effectively. In this paper, we compared the performance of a wide range of mainstream machine learning algorithms, aiming to identify the ones offering satisfactory detection performance and stability based on a large amount of ground truth data. With the goal of achieving real-time Twitter spam detection capability, we further evaluated the algorithms in terms of the scalability. The performance study evaluates the detection accuracy, the true/false positive rate and the F-measure; the stability examines how stable the algorithms perform using randomly selected training samples of different sizes. The scalability aims to better understand the impact of the parallel computing environment on the reduction of the training/testing time of machine learning algorithms.