Master Degree / Yüksek Lisans Tezleri
Permanent URI for this collectionhttps://hdl.handle.net/11147/3008
Browse
3 results
Search Results
Master Thesis Analysis of Feature Pattern Mining Approaches on Social Network: a Case Study on Facebook(Izmir Institute of Technology, 2017) Öztürk, Elif; Şahin, SerapPattern mining algorithms obtain patterns frequently seen in a database and complex graphs which are available from gene networks to social networks. Complex graphs contain lots of valuable information on their nodes or edges. For this reason, pattern mining algorithms can be used to extract data from complex networks. However, these algorithms usually work on the graphs whose nodes have a single label. If these algorithms are implemented on multi labeled (multi-attributed) complex graphs, their complexities belong to NP-Complete. For this reason, in this study, different approaches have been evaluated to find patterns. The goal is to understand related methods and algorithms with their pros and cons to obtain common feature patterns from multi-attributed complex graphs. We also selected Facebook social network complex graph data set (SNAP - Stanford University FaceBook anonymized data set) as an application domain and we analyzed the most frequent feature patterns on friendship relations.Master Thesis An Analysis of Information Spreading and Privacy Issues on Social Networks(Izmir Institute of Technology, 2017) Sayin, Burcu; Şahin, SerapWith Social Networks (SNs), being populated by a still increasing number of people, who take advantage of the communication and collaboration capabilities that they offer, density of the information, spread over SNs is increasing steadily. Furthermore, the probability of exposure of someone’s personal moments to a wider than expected crowd is also increasing. Hence, analyzing the spreading area and privacy level of any information through a SN is an important issue in social network analysis. By studying the functionalities and characteristics that modern SNs offer, along with the people’s habits and common behavior in them, it is easy to understand that several privacy risks may exist, for many of which people may be unaware of. We address this issue, focusing on interactions with posts in a SN, using Facebook as the research domain. As a novelty, we propose an application tool which visualizes the effect of potential privacy risks in Facebook and provides users to control their privacy. The proposed (and simulated) tool allows a Post Owner to observe the spreading area of his/her post, depending on the selected privacy settings of this post. Moreover, it provides preliminary feedback for all the Facebook users that have interacted with this post, to make them aware of the possible privacy changes, aiming to give them a chance to protect the privacy of their interaction on this post by deleting it when such a privacy change takes place.Master Thesis Matching of Social Media Accounts by Using Public Information(Izmir Institute of Technology, 2016) Çetinkal, Yağız; Şahin, Serap; Şahin, SerapProtection of private information on social networks (SNs) has become a serious and important topic since social network sites became popular and widely adopted worldwide. Usually people want their personal information to be known only by a small group of people including close friends and families. But sometimes they willingly accept to give some particular information about themselves to individuals which are neither a friend nor an acquaintance. Each SN has different purposes and people subscribe many of them. However, public information available on these sites reveals many aspects of user’s identity. In this work, it is shown that public information can be used to detect the different accounts of the same individual. This study is performed on two popular social media sites: Twitter and Facebook. Public attributes of the profiles such as real name, user name and status updates (tweets and posts) are used for comparing profiles on two SNs. Different data mining algorithms are compared for matching profiles. Also relationship between text similarity and total term counts of status updates is analyzed. Results show that simple features like real names, user names and status updates have high similarity between the accounts of the same users and these features can be used to detect profiles of the same user on different SNs. Also the more status updates a user posts on Facebook the more he will likely be detected by the matching schema. Thus, public information can be exploited to pose a threat to the privacy of the people on the Internet.