تباه کننده های بازاریابی رسانه های اجتماعی؟ یادگیری افزایشی معنی شناسی پنهان برای شناسایی هرزنامه های اجتماعی Who are the spoilers in social media marketing? Incremental learning of latent semantics for social spam detection

1 Introduction Spam became prevalent in the late 1990s and early 2000s when email was considered to be the primary tool for information exchange among individuals and firms. With the introduction of email spam filters, spammers have started looking at other platforms for better payoffs. One of these ‘‘money-making’’ platforms for spammers are social media sites (SMSs) that play an increasingly important role in our daily lives [1]. Nowadays, online social media data exhibits the 4Vs characteristics that are often used to describe Big Data, namely volume, velocity, variety, and veracity [2]. In terms of volume, the number of active users on Facebook and Twitter has reached, respectively, 1.55 and 0.32 billion in November 2015.1 There are over 500 million tweets generated on a daily basis.2 Besides signifying its importance in our daily lives, the features of online social media (i.e., creation and exchange of user-generated content, support for collective actions, and facilitation of diverse social interactions) denote its indispensable function of being a business tool for promoting e-commerce products and services. Thus, increasingly more e-commerce retailers choose online social media as a main marketing platform or a new ‘‘social CRM’’ tool that fosters instant interactions with potential consumers. SMSs likewise provide spammers with unprecedented opportunities to launch various attacks. Spammers perform deceptive acts [3, 4], conduct unfair trading activities [5, 6], and even make illegal profits [7] by posting social spam on SMSs. Social spam refers to low-quality information for which users do not ask or specifically subscribe to [8]. Social spam is used to launch phishing attacks [9], promote adverse websites [10], distribute malwares [11], and spread adverse messages [12, 13]. Embedded URLs in social spam direct users to adverts, malware, or pornographic websites (see Fig. 1). According to Nexgate’s state of social media spam report, there has been a 355 % growth of social spam in the first half of 2013 [14]. As well it has been revealed that more spammers were found on Facebook and YouTube than any other SMS. Grier et al. [15] reported that 8 % of the 25 million URLs posted on Twitter were phishes, malware, and scams.