In recent years, a considerable amount of research has been done on various topics in online social networks such as Facebook and Flickr. The fast increase in the popularity of these networks has prompted researchers to study how social networks form and evolve over time. This part of our research activity is devoted to investigate the social phenomena on networks. Here we study how the complex structure of social interactions, affects the dynamics of collective behavior such as the information propagation, the cooperation between selfish individuals, and the emergence of synchronization between interacting agents. Other topics of research concern the study of resilience of network, the detection of community structures, and the efficient search methods in social networks.
Signed Network Analysis
The study of social networks has attracted a great attention in recent years. Many traditional studies considered the networks as homogeneous binary networks, but indeed the relations can be different in many aspects. Signed networks are an example of these networks. In signed networks the links can have a sign the indicates whether the link is a friendship (trust) or enemy (distrust) relation. Indeed in these networks we have another source of information that can help us in the analysis of the network.
The works who study the signed networks can be categorized into three categories.
- The first category, which contains the first attempts in the study of signed networks, tries to study the structural features of signed networks. The main purpose of these studies, is to assess the amount of presence of social theories such as structural balance and status theory in real networks.
- The second group contains the studies aimed at sign prediction. These studies tries to predict the unseen signs in the network. The problem at hand can be formulated as a classification problem.
- Finally, the third group tries to detect the communities of in signed networks. The definition of community for signed networks differs dramatically from its definition for unsigned networks. The traditional definition of community for unsigned networks, contains the groups of densely connected nodes with sparse inter group links. In the signed networks the community is defined as a group of nodes with positive links between them, while the negative links appear between groups. This difference makes the unsigned networks community detection methods inappropriate for signed networks.
In this research we will focus on the three mentioned areas for the analysis of signed networks.
People involved: Erfan Tavakkoli, Mohammadreza Karimi, Nazgol Tavabi, Soheil Arabzade, Abbas Heseini, Ali Zarezade, Ali Khodadadi, Hamid R. Rabiee
Peer-to-Peer Social Networking
A Peer-to-Peer (P2P) social network consists of social communities which are constructed on top of the underlying P2P overlay network. In order to build a social community, each peer’s characteristic is gained by comparing with another peer’s behavior and then the similar degree of any two peers is computed by a predefined formula. Finally peers whose similar degrees are larger than the predefined similarity threshold will form a social community. If a peer can be put into several social communities, it can be allowed to join multiple communities or join the community with the largest similar degree. In this project we intend to build a P2P social network in which clients can join to the overlay via a laptop, cell phone, or PC, and share their data, picture, films,… with others. The architecture of the network has three main layers: DHT-based P2P Overlay responsible for constructing the overlay network, Plug-in layer responsible for creating the social network plug-ins, and application layer which employs the plug-ins and provides a user friendly interface for clients. In this network each peer joins the network by contacting with the bootstrap node to find its neighbors. After the joining process, the peer creates its own social community by adding its PC, cell phone, or any other devices. Through this P2P social network, the users are capable of finding professors or specific authors in a conference, access to the shared data even if the Internet connection isn’t available, or attend in some social activities such as instant messaging and photo sharing with friends.
People involved: Hoda Ayatollahi, Mohammad Farahani, Mohammad H. Eslamimehr, Mostafa Salehi, Hamid R. Rabiee
Mobile Social Networks
When the rapid evolution of mobile devices, the massive adoption of cell phones by the population and the important growth in social networks came together, the mobile social network was born. This topic goes through the migration of social networks to mobile devices and the new features that they could provide over their network infrastructure, specially cellular networks, and with the context awareness abilities and also, it covers some architectural notes for the mobile social networks and their limitations.
Here our focus on transmitting information (communicating) using a mixture of voice and data devices over networks including cellular technology and elements of private and public IP infrastructure (such as the Internet). ‘Mobile Social Networking’ (MSN) refers to all of the enabling elements necessary for the contribution (‘posting’ and uploading) and consumption (viewing/experiencing) of social media across a mobile network. If the user accesses a community service platform by way of any device that uses a cellular network, alone or in combination with a commercially-accessible wireless network that has access to cellular network operator-owned resources, then that activity is included in the scope of this research. Furthermore, mobile community operators and participants are, and can be, influenced by the platforms, trends and members of communities on the Internet.
People involved: Mohammah M. Aghajani, Mostafa Salehi, Hamid R. Rabiee
Online Social Networks Analysis
An important principle to start a project is that you must be aware of failed and success points of same projects. So because of our deciding to implement a specific-target social network, we begin gathering everything about famous social networks such as Facebook, Youtube, Yahoo! 360, Yahoo! Pulse, and etc. Now this part hasn’t been finished yet but we get a lot of information about failed and success reasons of social networks, how to start and develop a social network, and we can create a data bank statistics about social networks.
People involved: Mohammad H. Eslamimehr, Leila Talebpoor, Mohammah M. Aghajani, Amrollah Seifoddini, Mostafa Salehi, Hamid R. Rabiee
An Overview on Testbeds for Internet Services
Nowadays, many real world Internet services have various applications such as facilitating
communication between individuals. Mostly, these services are developed using Peer-to-Peer and
Cloud computing on wired and wireless infrastructures. For reducing costs and challenges of
development, the output quality of these services must be measured.
One specific tool for testing Internet services that has received prominent interest in recent years is
network testbed. Networks testbeds are one of the network infrastructures that play an important
role in increasing the quality of Internet services and providing better performance evaluation
results for them.
In this context, many network testbeds such as PlanetLab, Orbit Testbed, and Open Cirrus
Project were developed in research institutions and universities of various sizes. For example,
PlanetLab is a global research network that supports the development of new network services.
Since the beginning of 2003, more than 1,000 reserchers at top academic institutions and
industrial research labs have used PlanetLab to develpe new technologies for distributed
systems, peer-to-peer systems. This testbed currently have about 1,000 nodes at 545 sites.
Moreover, PlanetLab is supported by well-known computer companies like Intel and Cisco.
In this project, some existing testbeds in the areas of wired networks, wireless networks, and cloud
computing will be studied. In addition, we discuss architecures, design goals, challenges and
results of deploying such a testbeds from an analytical viewpoint. Moreoever, we propose some
detailed software architecture ideas for developing and customization of network testbeds in Iran.
Mehrdad Moradi, Mostafa Salehi, Hamid R. Rabiee
Prediction in Signed Social Networks
One of the questions in social networks studying is influences that users receive from others in social media sites. In recent years, different models is presented to analyze the relations between individuals and user influence to each other. On the other hand relations between users on social media sites often reflect a mixture of positive (friendly) and negative(antagonistic or opposition) interactions. For modeling these relationships we can used signed networks and their theories. This type of networks are displayed by labeled graphs: the nodes display users and the edges display relations among users. Positive label over an edge display a friendly relationship between the nodes connected and negative label display an antagonistic relationship.But the key question in this regard is: how the sign of an edge can affect on relations between its neighbors and the more general look at the whole of the entire network?. Or how the interaction between positive and negative relationships can affect the structure of social networks ? One of answers to these questions is prediction of the link sign between nodes: let we have a social network with signs on all of the edges except one of them, how we can determine sign of this edge with other information in the network? In this project we first formulating prediction of the link sign problem, Then investigate and present efficient mechanism for this problem. Finally, the efficiency of these methods evaluated through simulating them on real social networks. What we presented at our last work is a model of network creation game for evolution of signed networks based on structural balance theory. The players are nodes and their strategy is creating positive/negative/natural edges to all other nodes. the edges are undirected and when they’re created they represent bilateral friendship/antagonism/natural relationship between players, independently of which node created it. The sum of these edges is the resulting signed graph.
People involved: Mohammad Malekzadeh, Mostafa Salehi, Hamid R. Rabiee
Statistics and Characterization of YouTube Social Network
Many real world structures such as transportation channels, members of deviant groups, communication networks, cancer cells and social networks appear in form of dynamic networks, where interaction between nodes which represent objects in the environment results in a dynamic structure of the whole. Complex networks are being developed in order to analyze and design of these concepts. One specific area that has received prominent interest in recent years is the modeling of complex networks evolution. The way a network evolves plays an important role in analyzing its structure and behavior in future. Besides, it helps developers to take the necessary actions. Nodes enter the network, establish some links to others, and may decide to leave the gathering. Therefore we need a model to meet the growing needs of dynamic nature of such networks. Specific complex networks have specific attributes and features and so must our model. There are two approaches to follow. In the first approach, the model defines the behavior of its nodes such that one (or more than one) specific attribute (such as power-law degree distributions of nodes) can be satisfied. In the second approach we try to extract the nodes behavior such as network joining and leaving processes and link formation by studying and analyzing the desired network data. These models can be used to develop an appropriate algorithm which generates the observed data. In this project, some existing evaluation models of complex networks will be studied. In addition, extracted data from video-sharing site YouTube is analyzed and one of the most important statistical features, degree distribution, is discussed. Then by studying users activity over time, we show that in its early stages of evolution, the network don’t behave according to proposed models which are for next stages. Moreover, we observe that memberships of a user’s friends in a specified group have a great influence on his/her membership. But with increasing the number of these friends, his/her interest to join the group will be diminished.
In this project, three phases were followed: I) The first step was to study and investigate different growth models of complex networks. II) The second phase was to implement a crawler for YouTube video sharing site. We crawled more than 20000 swiss members of Youtube. The crawled data includes their uploaded videos, their comments, their friendship links and their subscriptions to YouTube channels. III) in the last step, we focused on patterns which were observed in the crawled data. Degree distribution of nodes is firstly studied. Afterwards we observed that members activity during their lifetime follows a specified pattern. Finally the effect of friendship on joining to channels were studied.
People involved: Erfan Zamanian, Mostafa Salehi, Hamid R. Rabiee