Summary by Adnan World is full of both good and bad human beings, depends how they behave while using any system. When I talking about Social network system users then we talk about two main types real user¨ªs or fake user¨ªs. Online Social Networks are targeted by the hackers today, since these networks are most widely used online systems, like Facebook. According to a recent research by Facebook, up to 83 million of its users are maybe fake, this figure is higher than their previous research figure: 54 Million. There are various techniques to find these fake users but majority of these techniques relay on assumption that Sybil (Fake user) accounts have difficulty friending legitimate users, and so tend to form their own communities, making them visible to community detection techniques to the social graph. Unfortunately success of these systems decreases as the Sybil adopts better ways to avoid the system trap. First noticeable thing is such users are not careful about the friendship requests. They just accept all of the friend requests. Hence this makes allot of Sybil friends with other Sybil¨ªs and hence a community of Sybil is formed. These friend loops is not limited to Sybil¨ªs but to real accounts as well. Sybil accounts are using advance techniques for example by hiring real humans to customize their profiles. Crowdsourcing is a tool to detect Sybil accounts. Designing a successful crowdsourced Sybil detection system requires that first answer fundamental questions like Accuracy of detecting the fake accounts? , Accuracy and Language and cultural barrier ? How much it would cost to check suspicious profiles? , Scalability of such a system with respect to millions of user¨ªs ?. The paper considers all these questions and finally proposes a scalable crowdsourced Sybil detection system based on the results. It uses trace driven data to achieve both accuracy and scalability with reasonable costs. Summary by Zahid Social Turing Tests: Crowdsourcing Sybil Detection Miriam Metzger, Haitao Zheng and Ben Y. Zhao Summary When number of internet users getting increased than there are lot of hackers introduced to hack the systems or data. In the world of internet there are different tools those spread spams and malware like Sybil¡¯s, these tools create lot of spams and malware into Online Social Networks. To detect such spams and malware authors proposed one system called crowedsourced Sybil detection system for Online Social Networks. They conduct large number of user study to detect the Sybil accounts from Facebook and Renner Networks. During the experiments authors detect various conditions of ¡°experts¡± and ¡°turkers¡±. And they get the results from these conditions and they used all these ways to detect the spams and malwares. They used these result to derive the design of multi-tire crowdsourced Sybil detection system. This shows high scalability performance on both ways, like standalone system or as a complimentary technique to current tools. Summary by Patrick This paper presents a system for crowdsourcing Sybil detections. A Sybil is a fake user account on a social network. Mainly used for sending spam targeted at other users. Mechanical Turks are human hired over the internet to do work that is above what machines can do. I can be transcribing audio, video or writing papers. The authors perform an experiment where they attempt to use mechanical turks to determine fake user accounts (Sybil accounts) by having them rate the users profile. They collect data from Facebook and Renren, classifying sets of accounts and having experts and turks rate them. Their results show that experts (People in the field of computer science) are more adept at determining a Sybil account over turks. They suggest a system for partially-automated detection of Sybil accounts. Summary by Ioana Social Turing Tests: Crowd sourcing Sybil Detection -Gang Wang, Manish Mohanlal, Christo Wilson, Xiao Wang, Miriam J. Metzger, Haitao Zheng, Ben Y. Zhao. The paper introduces a crowd sourced detection system for Sybil accounts. These accounts represent fake identities that are created by malicious users, and used in spam and malware. The motivation of this work is represented by the way in which the existence of such accounts affects the online social networks. In the first part of the paper, the authors investigate whether building such a system is feasible or not. More precisely, they are interested in whether Sybil accounts can be accurately detected by users or not, whether different demographic aspects affect the detection process, if the fatigue plays an important role in the detection and how cost effective would be such a system. After investigating all the above aspects, the authors propose a practical system and validate it. The ground-truth data sets of users profiles employed in the paper are obtained from Renren, Facebook India and Facebook US. In order to establish how different users can detect Sybil account the authors conduct a user study. The categories of users used are: - Experts( represented by computer science professors and graduate students ) - Turks( from crowd sourcing websites ) - Sociology undergraduates ; The main finding is the fact that humans can identify between Sybil accounts and legitimate users, but using the majority opinion of a crowd gives more accurate results. The practical aproach of the authors is a two layered system ( filtering and crowdsourcing layer), and the main aspects considered in designing it are scalability, accuracy and privacy. In the validation part, it is showed that the proposed solution produces less that 1% in false positives and negative, and it is cost efficient. Questions : 1. How is the reliability of the users discussed in the paper? 2. What other factors can influence the detection accuracy ( discussed in the user study )? 3. How would you extend the system in order to provide user privacy ?