Spam detection in online social networks by deep learning

Abstract

Twitter spam is one of the most important problems that professionals have to deal with in social networks on the internet. For this problem, the researchers presented some solutions, mostly based on a number of different methods considering learning. The means and techniques used at the current time has achieved a good ratio of the accuracy based on the so-called methods of blacklisting in order to determine the undesirable activities in relation to send and receive an e-mail on social networks that based on the conclusions obtained from previous experiments and studies. However, methods that rely on automated learning are not capable of detecting spam activities in proportion to the real scenarios. We see that methods called blacklist methods are not able to meet the disparities we see in activities related to the transmission of such a message, because manually checking Unique Resources Locaters (URLs) is more time-consuming task. In this study, we present a deep learning method for spam detection in witter. For this purpose, the Word2Vec based on representation is first trained. Then we use binary classification methods to distinguish the spam and the nonspam tweets. The empirical results conducted on tweets prove that the selected methods outperform the classical approaches.