Andrea De Salve

The impact of user's availability on On-line Ego Networks

We have defined and implemented a Facebook application to log a Facebook dataset.We have studied and validated the structural properties of the whole dataset and of the Dunbar ego networks.We have analyzed the interactions of the users.The availability of the users in the Dunbar ego networks have been investigated.Our results reveal the presence of the temporal homophily property in the Dunbar ego networks. Online Social Networks (OSNs) are the most popular applications in todays Internet and they have changed the way people interact with each other. Understanding the structural properties of OSNs and, in particular, how users behave when they connect to OSNs is crucial for designing user-centered systems. Results about OSNs demonstrated that the relationships that an individual (ego) maintains with other people (alters) can be organized into a set of circles (named Dunbars circles) according to the ego network model. The study of the impact of ego networks structure on the availability patterns of users is seriously limited by the lack of information about users availability patterns. In this work we contribute to fill this gap by analysing availability information of a sample of Facebook users. The data reveal a number of strong temporal dependencies (or temporal homophily) which provide insights into the availability pattern that characterize an ego network.

DiDuSoNet: A P2P architecture for distributed Dunbar-based social networks

Online Social Networks (OSNs) are becoming more and more popular on the Web. Distributed Online Social Networks (DOSNs) are OSNs which do not exploit a central server for storing users data and enable users to have more control on their profile content, ensuring a higher level of privacy. In a DOSN there are some technical challenges to face. One of the most important challenges is the data availability problem when a user is offline. In this paper we propose DiDuSoNet, a novel P2P Distributed Online Social Network where users can exercise full access control on their data. Our system exploits trust relationships for providing a set of important social services, such as trustness, information diffusion, and data availability. In this paper we show how our system manages the problem of data availability by proposing a new P2P dynamic trusted storage approach. By following the Dunbar concept, our system stores the data of a user only on a restricted number of friends which have regular contacts with him/her. Differently from other approaches, nodes chosen to keep data replicas are not statically defined but dynamically change according to users churn. In according to our previous work, we use only two online profile replicas at time. By using real Facebook data traces we prove that our approach offers high availability.

Trusted Dynamic Storage for Dunbar-Based P2P Online Social Networks

Online Social Networks (OSNs) are becoming more and more popular in today’s Internet. Distributed Online Social Networks (DOSNs), are OSNs which do not exploit a central server for storing users’ data and enable users to have more control on their profile content, ensuring a higher level of privacy. The main challenge of DOSNs comes from guaranteeing availability of the data when the data owner is offline. In this paper we propose a new P2P dynamic approach to the problem of data persistence in DOSNs. By following Dunbar’s approach, our system stores the data of a user only on a restricted number of friends which have regular contacts with him/her. Users in this set are chosen by considering several criteria targeting different goals. Differently from other approaches, nodes chosen to keep data replicas are not statically defined but dynamically change according to users’ churn. Our dynamic friend selection achieves availability higher than 90% with a maximum of 2 online profile replicas at a time for users with at least 40 friends. By using real Facebook data traces we prove that our approach offers high availability even when the online time of users is low.

Epidemic Diffusion of Social Updates in Dunbar-Based DOSN

Distributed Online Social Networks DOSNs do not rely on a central repository for storing social data so that the users can keep control of their private data and do not depend on the social network provider. The ego network, i.e. the network made up of an individual, the ego, along with all the social ties she has with other people, the alters, may be exploited to define distributed social overlays and dissemination protocols. In this paper we propose a new epidemic protocol able to spread social updates in Dunbar-based DOSN overlays where the links between nodes are defined by considering the social interactions between users. Our approach is based on the notion of Weighted Ego Betweenness Centrality WEBC which is an egocentric social measure approximating the Betweenness Centrality. The computation of the WEBC exploits a weighted graph where the weights correspond to the tie strengths between the users so that nodes having a higher number of interactions are characterized by a higher value of the WEBC. A set of experimental results proving the effectiveness of our approach is presented.

A Privacy-Aware Framework for Decentralized Online Social Networks

Online social networks based on a single service provider suffer several drawbacks, first of all the privacy issues arising from the delegation of user data to a single entity. Distributed online social networks (DOSN) have been recently proposed as an alternative solution allowing users to keep control of their private data. However, the lack of a centralized entity introduces new problems, like the need of defining proper privacy policies for data access and of guaranteeing the availability of user’s data when the user disconnects from the social network. This paper introduces a privacy-aware support for DOSN enabling users to define a set of privacy policies which describe who is entitled to access the data in their social profile. These policies are exploited by the DOSN support to decide the re-allocation of the profile when the user disconnects from the social network. The proposed approach is validated through a set of simulations performed on real traces logged from Facebook.

Distributed Coverage of Ego Networks in F2F Online Social Networks

Although most online social networks rely on a centralized infrastructure, several proposals of Distributed Online Social Networks (DOSNs) have been recently presented. Since in DOSNs user profiles are stored on the peers of the users belonging to the network, one of the main challenges comes from guaranteeing the profile availability when the owner of the data is not online. In this paper, we propose a DOSN based on a friend-to-friend P2P overlay where the users data is stored only on friend peers. Our approach is based on the ego-network concept, which models the social network from the local point of view of a single user. We propose a distributed algorithm which is based on the notion of coverage of the ego-network, assures that users store their data only on the peers of their friends,, that each online user can retrieve the private data of its offline friends through a common online friend. We formalize this as a Neighbour Dominating Set problem. A set of experimental results conducted on real Facebook dataset show the effectiveness of our approach.

Evaluation of Structural and Temporal Properties of Ego Networks for Data Availability in DOSNs

The large diffusion of Online Social Networks (OSNs) has influenced the way people interact with each other. OSNs present several drawbacks, one of the most important is the problem of privacy disclosures. Distributed Online Social Networks (DOSNs) have been proposed as a valid alternative solution to solve this problem. DOSNs are Online Social Networks implemented on a distributed platform, such as a P2P system or a mobile network. However, the decentralization of the control presents several challenges, one of the main ones is guaranteeing data availability without relying on a central server. To this aim, users’ data allocation strategies have to be defined and this requires the knowledge of both structural and temporal characteristics of ego networks which is a difficult task due to the lack of real datasets limiting the research in this field. The goal of this paper is the study of the behaviour of users in a real social network in order to define proper strategies to allocate the users’ data on the DOSN nodes. In particular, we present an analysis of the temporal affinity and the structure of communities and their evolution over the time by using a real Facebook dataset.

Logical key hierarchy for groups management in Distributed Online Social Network

Distributed Online Social Networks (DOSNs) have recently been proposed to shift the control over user data from a unique entity to the users of the DOSN themselves. In this paper, we focus our attention on the problem of privacy preserving content sharing to a large group of users of the DOSNs. Several solutions, based on cryptographic techniques, have been recently proposed. The main challenge here is the definition of a scalable and decentralized approach that: i) minimizes the re-encryption of the contents published in a group when the composition of the group changes and ii) enables a fast distribution of the cryptographic keys to all the members (n) of a group, each time a new user is added or removed from the group by the group owner. Our solution achieves the above goals, providing performance unattained by our competitors. In particular, our approach requires only O(d·logn) encryption operations when the group membership changes (eviction), and only O(2·logn) when a join occurs (where d is an input parameter of the system). The effectiveness of our approach is evaluated by an experimental campaign carried out over a set of traces from a real online social network.

Privacy and temporal aware allocation of data in Decentralized Online Social Networks

Distributed Online Social Networks (DOSNs) have recently been proposed to grant users more control over the data they share with the other users. Indeed, in contrast to centralized Online Social Networks (such as Facebook), DOSNs are not based on centralized storage services, because the contents shared by the users are stored on the devices of the users themselves. One of the main challenges in a DOSN comes from guaranteeing availability of the users’ contents when the data owner disconnects from the network. In this paper, we focus our attention on data availability by proposing a distributed allocation strategy which takes into account both the privacy policies defined on the contents and the availability patterns (online/offline) of the users in order to allocate their contents on trusted nodes. A linear predictor is used to model and to predict the availability status of the users in a future time interval, on the basis of their past temporal behaviour. We conduct a set of experiments on a set of traces taken from Facebook. The results prove the effectiveness of our approach by showing high availability of users’ profiles.

Predicting the availability of users' devices in decentralized online social networks: Predicting the availability of users devices

The understanding of the user temporal behavior is a crucial aspect for all those systems that rely on user resources for daily operations, such as decentralized online social networks (DOSNs). Indeed, DOSNs exploit the devices of their users to take on and share the tasks needed to provide services such as storing the published data. In the last years, the increasing popularity of DOSN services has changed the way of how people interact with each other by enabling users to connect to these services at any time by using their personal devices (such as notebooks or smartphones). As a result, the availability of data in these systems is strongly affected (or reflected) by the temporal behavior of their users in terms of connections to DOSNs. In this paper, we propose the use of linear predictors to address the problem of the availability of user devices and, hence, data in DOSNs. To validate the proposed approaches, we evaluated their performance conducting a set of simulations exploiting a dataset of temporal information concerning the connections to Facebook collected from a set of users.