微博内作弊和推广联盟的检测算法研究

发布时间：2018-05-16 04:26

本文选题：微博 + 推广联盟　；参考：《大连理工大学》2013年硕士论文

【摘要】：近几年来,微博类型的社交网络作为新兴的网上交流平台,已经受到大众的强烈欢迎。然而这类平台在给大家带来便利的同时,也给营销者和作弊者提供了传播信息的平台。不少的作弊账户开始在社交网络上开始作弊行为,而且据有效统计,社交网络上的作弊成功率远高于邮件和搜索引擎作弊行为。据研究发现,在微博上已经存在大量的推广联盟和作弊联盟,这样的联盟比传统的作弊方式,比如垃圾邮件方式,或者单一的推广或者作弊账户的危害大的多,严重影响了微博的安全环境。不论是推广联盟还是作弊联盟,他们都控制大量的微博账户来帮助他们传播对正常用户来说无用的或者有害的信息。由于掌控着大量的账户,信息的传播范围就会比掌控单一账户来传播的范围要大得多,因此从推广或者作弊联盟中传播出来的信息会影响更多的正常用户。由于传统的检测方法主要是检查单个的账户或者单条微博,而不是将整个的联盟查找出来,因此,急需一种方法来将微博内的推广和作弊联盟检测出来,以保证微博平台的可靠性和安全性。在这篇文章中,研究了很多的已有的工作,提出了一种可靠的框架来同时检测作弊和推广联盟。这个框架主要由三部分组成,首先将那些为了同样目的而发布URL的账户连接起来组成一个用户图,然后从这个用户图中提取可能是为了作弊或者推广目的而存在的候选联盟,最后提取联盟的特征来区分它们的目的。这个框架中最重要的一部分就是一种由链接驱动的估计算法,来估计微博账户发布链接的目的之间的相似度。另-个重要的贡献是为区分联盟目的而提出了很多有用的特征,然后用可靠的分类算法来对他们的目的进行分类。最终在大规模的微博数据集上的实验结果表明,这个框架对检测微博内推广和作弊联盟有较高的准确度,而且能准确地区分它们的实际目的。
[Abstract]:In recent years, Weibo type social networks as a new online communication platform, has been strongly welcomed by the public. However, such platforms not only provide convenience, but also provide a platform for marketers and cheaters to disseminate information. Many cheating accounts begin to cheat on social networks, and according to effective statistics, the success rate of cheating on social networks is much higher than that on email and search engines. According to research, there are already a large number of promotion alliances and cheating coalitions on Weibo, which are far more harmful than traditional cheating methods, such as spam, or a single promotion or cheating account. The security environment of Weibo is seriously affected. Both promotion and cheating coalitions control a large number of Weibo accounts to help them spread information that is useless or harmful to normal users. With a large number of accounts under control, information can spread much more widely than a single account, so information from promotion or cheating coalitions can affect more regular users. Because the traditional detection method is mainly to check a single account or a single Weibo, rather than find out the whole alliance, so it is urgent to use a method to detect the promotion and cheating alliance in Weibo. To ensure the reliability and security of Weibo platform. In this paper, we study a lot of existing work and propose a reliable framework to detect cheating and promote coalitions at the same time. The framework consists of three parts, which first connect accounts that publish URL for the same purpose to form a user graph, and then extract candidate coalitions that may exist for cheating or promotion purposes. Finally, the features of the alliance are extracted to distinguish their purpose. The most important part of this framework is a link-driven estimation algorithm to estimate the similarity between the purposes of Weibo account publishing links. Another important contribution is that many useful features are proposed to distinguish the purpose of a coalition, and then their purposes are classified by a reliable classification algorithm. Finally, the experimental results on large scale Weibo datasets show that the framework has high accuracy in detecting Weibo extension and cheating coalitions, and it can accurately distinguish their actual purpose.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP393.092

【相似文献】