=Paper=
{{Paper
|id=None
|storemode=property
|title=Software Agents for Distributed Social Networking
|pdfUrl=https://ceur-ws.org/Vol-892/paper4.pdf
|volume=Vol-892
|dblpUrl=https://dblp.org/rec/conf/woa/FranchiT12
}}
==Software Agents for Distributed Social Networking==
https://ceur-ws.org/Vol-892/paper4.pdf
Software Agents for Distributed Social Networking
Enrico Franchi, Michele Tomaiuolo
Dipartimento di Ingegneria dell'Informazione
Università di Parma
Parma, Italy
{efranchi, tomamic}@ce.unipr.it
Abstract—Especially in the case of completely distributed or amount of resources needed. Moreover, regarding censorship
federated social networking platforms, multi-agent systems can issues, a P2P system essentially solves them by design. Without
play an important role. In particular, multi-agent systems have a central entity, nobody is in the position of censoring any data
been used as (i) an underlying layer or a middleware for nor may be held legally responsible for the diffusion of
developing social networking platforms, (ii) a technology to censurable data: the sole owners and responsible of the data are
increase the autonomous and intelligent behaviour of existing the users themselves.
systems and (iii) a tool to develop simulation environments for
studying both online and offline human social networks. In this Especially in the case of completely distributed or federated
paper we propose the integration of multi-agent technology into social networking platforms, multi-agent systems can play an
Blogracy, a novel peer-to-peer, anonymous and uncensurable important role. In fact, one of the very specific features of
social networking platform. The resulting system augments the multi-agent systems is the sociality of agents, i.e. their ability
platform with locality and proximity groups, making it fit for to communicate in a semantic way and develop trust
pervasive computing scenarios, exploiting the adaptivity, relationships among them. Moreover, agents can express their
proactivity and negotiation ability of multi-agent systems. communication acts by means of acknowledged standards, like
FIPA, for interoperability among diverse systems, and
Distributed Micro-blogging, Peer-to-peer Computing, Social exchange messages directly, in a peer-to-peer way. So, it is not
Networking, Multi-Agent Systems surprising that these two technologies are often applied
together for developing advanced social platforms. In
particular, multi-agent systems have been used as (i) an
I. INTRODUCTION underlying layer or a middleware for developing social
Users of popular social networking sites are becoming networking platforms, (ii) a technology to increase the
increasingly wary of the privacy issues they face. Often, the autonomous and intelligent behaviour of existing systems and
perceived problems are related to the possible leakage of (iii) a tool to develop simulation environments for studying
personal information to extraneous persons, or to workmates. both online and offline human social networks.
However, privacy threats can also come from the service
providers which, being mostly centralized systems, maintain For the first type of solution, many of the distinguishing
full access and control over published data. features of multi-agent systems can be fully exploited. In fact,
multi-agent systems provide semantic communication among
From the technical point of view, scaling centralized agents, which is handy for expressing all the different actions
systems to tens or hundreds of million of users is a hard that users can perform on a social platform. The different
challenge, which can be faced if enough resources are performatives of messages can be understood according to their
deployed. Most companies rely on mining users' data for pragmatics meaning, and applied according to existing trust
supporting targeted advertisement. This behavior poses serious relations among the users and their respective agents. Also,
threats to privacy and data protection issues. Quite complex negotiation protocols can help creating
consequently, social networking sites guide their users into acknowledgements and trust among users, in an automatic or
“walled gardens”, without giving users full control over their assisted way, without exposing sensitive data. Mobility can
own information because such information constitutes much of also be useful for moving the computation closer to data, if
their company value [1]. Moreover, service providers are in the massive analysis has to be performed, but can also be handy for
position to effectively perform a-priori or a-posteriori adding functionality to a node of a distributed social platform
censorship, or to disclose all the information they have, no or to a user's client application.
matter how private, to other entities. They can perform such
actions either motivated by selfish interests or forced under In the second case, agents are mainly exploited because of
legal terms and other forms of pressure. their proactive and reactive behaviours, for providing
recommendations of both users and content and for providing
On the other hand, P2P systems essentially achieve personalization of results. Reactive abilities fit particularly well
automatic resource scalability, in the sense that the availability into a social networking environment, where events happen
of resources is proportional to the number of users. This continuously and users can be easily distracted by the huge
property is especially desirable for media sharing social information overflow which is associate with richly
networking systems, considering the exceptionally high interconnected social networks. Sensing the environment and
executing automatic tasks can reduce this overload Not many existing social networking platforms are based
significantly. Goal-oriented behaviours, on the other hand, can entirely on multi-agent systems. Among the research works,
support users in persecuting their long term objectives about MAgNet [7] is a multi-agent system built using JADE [8] and
friend and content discovery, i.e. finding known persons FOAF [9]. It is a prototype application providing social
registered in the network, making new acquaintances with oriented services to mobile users, i.e. defining groups of users
users with common interests, finding interesting content from and arranging group events. In [10], authors discuss the
new sources or hidden among other less relevant data. advantages of using multi-agent technologies for building
social platforms. They also underline some existing issues,
Finally, multi-agent systems are a powerful tool for mainly in terms of overlay infrastructure, navigability of the
simulating the behaviour of online social networks, in the same social network, existence of specific ontologies.
way they have been used for simulating the behaviour of
persons in real social environments for a long time. In fact, A larger number of systems exploit multi-agent technology
multi-agent systems have proved to be very effective in the for augmenting existing social platforms. For example, in [11],
simulation of social networks, both during their initial creation an agent-based photo searching and recommender system for
and development and during their further operation. They flickr.com is proposed. In [12], authors propose an approach
allow to describe the behaviors of individuals, mimicking the for finding an expert in a social network. A user's profile is not
actions of human users in similar contexts, and to analyze the supposed to be completely available, and instead is learned by
associated emerging behaviour of the network as a whole. This an agent, by evaluating exchanged messages and the user's
way, multi-agent systems can provide precious insights for referrals. In [13], authors present a model of a recommender
further improvement of existing social platforms. system based on social networks, autonomous agents and trust
relationships. The aim is to both reach information not
II. RELATED WORK available in close nodes and filter information to be processed.
The system is analyzed with varying network density,
Various solutions are being proposed to overcome the preference heterogeneity and knowledge sparsenesss. In [14],
centralized architecture of the most widespread social the problem of automatic trust negotiation is contextualized to
networking platforms. Many of these proposals follow a multi-agent systems. Agents are used to negotiate and build
federated approach, allowing users registered on a certain trust among users, disclosing data and privacy policies
server to create relationships with users of other servers. incrementally and reciprocally. This is especially useful for
Others are full-fledged peer-to-peer systems, usually based on connecting users in a social network, disclosing only the
a distributed hash table (DHT). minimal possible set of profile attributes.
Federated social networking systems allow users registered Another widespread application of multi-agent systems in
on a certain server to create relationships with users of other the field of social networks is simulation. In [15], for example,
servers. The best known examples are Diaspora1 and a model of social network based on the notion of “circles” is
StatusNet2. Diaspora servers communicate by means of an ad- simulated over a multi-agent system. In [16], there is an
hoc federation protocol and the standard Salmon protocol3 for example of a simulation platform specifically designed for
comments. StatusNet (formerly known as Laconica) adheres to studying social networks. More in general, Ascape, NetLogo,
the OStatus standard protocol for the interconnection of MASON, Repast and Swarm are among the best known
various servers and uses a number of existing protocols for platforms for agent simulation, often used to study emerging
interoperability with other networks. behaviours and features of both online and offline social
networks.
Various social networking systems are being developed on
the basis of peer-to-peer communications and DHT indexing.
Among these, PeerSoN [2][3] is a prototype designed to III. RESILIENT MICROBLOGGING
provide encryption, decentralization and direct data exchange While many authors argue for the distribution and openness
in the field of social networks. A DHT is used to trace the of social networking and micro-blogging services, few usable
user's network presence and for obtaining the index of the implementations exist, either in the field of federated networks
user's recent content. LotusNet [4] is a model of a social or as fully distributed solutions. Considering the existing or
network to be built over Likir. Likir itself is a secured DHT, proposed solutions, we therefore present a new system, which
which requires a user to be authenticated according to an IBE we named Blogracy4. Essentially, it is an anonymous and
(Identity-Based Encryption) scheme, before participating in uncensurable microblogging platform, built incrementally over
the network. Safebook [5] is based on a DHT and a network of BitTorrent, a popular and resilient file-sharing service.
socially close peers, defined Matryoshka. Peers in a user's
Matryoshka are trusted and support the user by anonymizing The architecture of the application is modular and is build
communications and replicating content and profile around two basic components: (i) an underlying module for
information. Persona [6], though not being a distributed social basic file sharing and DHT operations, possibly exploiting an
network, uses an interesting Attribute-Based Encryption existing implementation, and (ii) an OpenSocial container, i.e.,
protocol for protecting access to users' content. It allows each a module providing the services of the social platform to the
user to assign credentials to various groups of “friends”, for local user, to be accessed through a web interface. Additionally,
accessing protected content. the system supports autonomous agents for providing (i)
recommendations of both users and content, (ii) personalization
of results, (iii) trust negotiation mechanisms. In the following
1 http://joindiaspora.com/
2 http://status.net/
3 http://www.salmon-protocol.org/ 4 http://www.blogracy.net/
paragraphs we will describe the most distinguishing features specification for the syndication of activities taken in social
realized in Blogracy over this extensible architecture. web applications and services. In Blogracy, the personal feed is
eventually signed to avoid tampering and then shared using the
underlying file-sharing platform.
Clearly, an application that does not provide explicit
representation for the user’s profile and contacts should not be
considered a social networking application. Essentially, in
Blogracy users define and manage a list of other users,
represented by their IDs. A user is not required to publish his
profile, nor the network of his social relations. However, if he
does, the profile, containing partial or full information, can be
retrieved as any other shared file and its magnet-uri can be also
reported in the user’s feed. In case privacy needs to be added,
cryptography shall be used. At the current stage, for exporting
profiles and contacts, Blogracy adopts Portable Contacts6 with
OpenSocial7 extensions, a format which has some benefits
from the interoperability point of view, being quite simple and
Figure 1. Blogracy architecture well supported by existing large social networks and mail
systems. It also allows to associate tags with each user, thus
matching the basic data structure managed by Blogracy.
For its basic operation, Blogracy exploits a peer-to-peer
file-sharing mechanism and two logically separated DHTs. One of the technical issues of a peer-to-peer microblogging
Users in Blogracy have a profile and a semantically application is data availability; in fact, popular content will
meaningful activity stream, which contains their actions in the quickly gain lots of seeds, while posts published by peripheral
system (e.g., add a post, tag a picture, comment a video). One users, with few contacts and sparse online presence, will
DHT maps the user’s identifier with his activity stream, which instead suffer poor availability to the extent that it is possible
also contains a reference to the user’s profile and references to that the publisher remains the only seed for his own new posts.
user generated content (e.g., posts, comments). These In some systems focused on distributed data storage, like
references are keys of the second DHT, which are then Freenet, the problem is addressed through multiple replication
resolved to the actual files. The files are delivered using the of all published resources. However, in modern peer-to-peer
underlying peer-to-peer file-sharing mechanism. networks, the hostile behavior of some nodes has to be taken
for granted; pollution and other kinds of attacks cannot be
Among the features of public online information systems, underestimated. What we foster, instead, is a replication system
and in particular in the case of micro-blogging and social based on acquaintances. Essentially, an introducing user is
networking applications, anonymity or pseudonymity are often responsible to introduce the invited as smoothly as possible.
a requirement. But, even under anonymity or pseudonymity, This kind of mechanisms is thoroughly analyzed in
users’ content need to be verified for authenticity and integrity. [19][20][21], with special regards to (i) content replication in
Blogracy uses a key-based identity scheme [17], where a user’s peer-to-peer storage and (ii) the problem of peers with low
public key is used directly to represent the user. This way, all availability in completely decentralized systems. In fact, using
content produced by the user can be easily verified against his some kind of fallback strategies for sharing non popular
public key, which is also his own main identifier. Moreover, for resources may improve the system performance regarding data
assuring anonymity at the lower network level, various availability.
anonymizing technologies exist, varying from simple proxies
to complex mix-net schemes, and can be integrated into the Another important issue is interoperability with other
platform. existing online social networking and micro-blogging
platforms. In principle, since Blogracy handles users’ feeds in
For publishing confidential information, accessible only to the form of Activity Streams, it can also manage similar feeds
a restricted circle of contacts, Blogracy supports attribute-based obtained in other ways, seamlessly integrating content from
encryption. Similarly to Persona, Blogracy privacy model uses web blogs and from the peer-to-peer network. Interoperability
attribute credentials for protecting access to sensible content, with more traditional news-feeds, web-based micro-blogging
creating a sort of very flexible “circles”, i.e., parametrized posts, and content distributed over the peer-to-peer network is
roles to be assigned to users for granting a certain set of access thus guaranteed, provided that the stream semantics is correct.
rights. The encryption scheme is based on the CP-ABE On the other hand, resources distributed through Blogracy can
protocol (Cyphertext-Policy Attribute-Based Encryption) [18]. be easily replicated over the web. Since the actual system
Once users can be distinguished by their ID, i.e., the hash architecture has a web interface, for user operation, it is
of their public key, it is also possible to associate additional relatively simple to host a Blogracy instance on a remote node
information with them, including personal profile and personal and configure it for public access, acting as a gateway for
activity stream. The activities of a user are represented as a Blogracy public content.
flow, which friends and followers are interested into and want Finally, apart from requesting updated feeds at startup,
to subscribe to. In Blogracy, personal activities are included followers should be timely notified that one of their followees
into a standard ActivityStream5 feed, which is eventually updated some resource. Traditionally the strategies are: (i) pull,
signed to avoid tampering. Activity Streams is an open format
6 http://portablecontacts.net/
5 http://activitystrea.ms/ 7 http://opensocial.org/
i.e., the observer periodically checks the observed resource for Over the Blogracy OpenSocial container, we are also
updates or (ii) push, i.e., the update is automatically announced integrating some functionalities for pervasive online social
to the observer. Apart from relying on the DHT, Blogracy networking, specifically for realizing locality and proximity
benefits from the peer-to-peer messaging facility provided by groups. For this purpose, each node of the social network will
the file-sharing protocol. In fact, for their basic operation, file- hosts multiple agents, with different levels of agency. Some of
sharing systems need to keep track of the peers that are the more important agents are (i) the Neighborhood Manager
currently seeding or downloading a certain file (sometimes agent (NM), which cooperates with lower level agents to
collectively defined as a “swarm”). So, advertising about a discover the users in its neighborhood; (ii) the Trust Negotiator
new feed is simply a matter of contacting the peers that are agent (TN), that is involved in the decisions regarding privacy
sharing the superseded version of the user's feed. and data access and (iii) the OpenSocial agent, that provides a
bridge towards the underlying Blogracy modules.
As described in the previous sections, Blogracy relies on
the BitTorrent protocol for basic file-sharing, and uses a DHT A user may own multiple nodes (e.g., an instance on the
mechanism for indexing the users' feeds. Specifically, we smart-phone and an instance on his home computer) and since
implemented the system exploiting Vuze, a popular BitTorrent the actual location of the user is important for our application,
client (formerly known as Azureus) implemented in Java and the nodes in the different device negotiate which should be
available as open source software. In particular, the specific considered active (i.e., which one determines the user
DHT of Vuze (known as DDB) has a set of generic primitive location): (i) the nodes determine which is the device that
queries that fit our purposes better than the Mainline DHT of registered an explicit user action or (ii) they ask the user to
other BitTorrent applications. Moreover, the Vuze platform has select the device he is currently using.
a modular architecture, where functionality can be added with
plug-ins. The main application exposes to the plug-ins only a Apart from the personal circles defined by each user, we
restricted interface, which is nonetheless sufficient for our also have two additional kinds of groups: (i) Proximity groups
purposes. and (ii) Location groups. Proximity groups are centered on
each member of the social networking system and represents
physical closeness to such member. Proximity groups are
IV. INTELLIGENT, PERVASIVE SOCIAL NETWORKING extremely fluid, in the sense that users can physically move
The Blogracy system itself relies only on users' nodes for and consequently the set of users belonging to a Proximity
its operation. Thus users need to perform background tasks on group varies in time. Each user configures the sticky-ness of
their own, in a distributed way. On the basis of the experience his Proximity group, i.e., how long the other users are
gained developing AOIS [22], we are integrating a layer of considered part of it after they are no longer physically close to
autonomous agents into the system, for assisting the user in him. Although a Proximity group may be entirely public, for
finding new interesting content and connections and for privacy reasons it is safer to consider only Proximity groups
pushing the local user's activities to followers. that are subset of other groups (or to the set union of all groups,
i.e., only “friends” are part of a Proximity group). The
In particular, a personal assistant (PA) monitors the local Neighborhood Manager agent informs the OpenSocial agent
user's actions in the platform and learns the user's profile, when users enter and leave the Proximity group and the latter
beyond information provided explicitly. The PA receives the notifies the OpenSocial container about it.
user’s queries, forwards them to the available information
finders (IF) and presents the results to the user. Moreover, a PA On the other hand, a Location group (i) is associated with
provides the local user with recommendations about possibly the users in the proximity of a given location (e.g., a classroom
interesting content and connections available in the network. or a museum room), (ii) has a host, i.e., a node that both
Another task performed by the PA is the personalization of identifies and supports the group and (iii) is associated with a
results. In fact, as a social network becomes larger and more location profile, which can be either hosted on the central
richly interconnected, users unavoidably face some form of server or on the device itself. In fact, a location, although
information overflow. A personal agent, on the basis of a user's logically different from a regular user, works in the same way
profile, can arrange presented data in a way to give evidence to and a Location group is essentially a Proximity group for the
the most interesting bits. location.
An Information Finder (IF) is an agent that searches A generic Trust Negotiation protocol may be needed since
information on the repository contained into the node where it users joining a proximity or location group are not necessarily
lives, on the basis of an automatic TF-IDF indexing and connected a priori in the social network, and they may need to
explicit hashtags associated with local posts. It provides this acknowledge their profile attributes before practical social
information both to its user and to other trusted users. An IF interaction. Such a negotiation requires the controlled exchange
receives users’ queries, finds appropriate results and filters of credentials and policies, without disclosing unnecessary
them on the basis of its user’s access policies. sensible information, yet establishing trust if possible. In [14]
we already presented a generic library supporting zero-
An information pusher (IP) is an agent that monitors the knowledge proof for attribute verification, which facilitates the
changes in the local repository and pushes the new information creation of trust in similar situations.
to the PA of interested subscribers who are currently connected.
The IP can forward content produced both by the local user or Agents present different degrees of autonomy and
by remote acquaintances to other contacts, according to privacy intelligence. For example, agents such as the lower level agents
preserving policies and to recent queries made by other users. are mostly reactive agents that inform the NM agent when a
new node is discovered. The NM agent itself has some degrees
of autonomy and intelligence: (i) it aggregates information
from the agents that discover new peers, (ii) informs the [4] L.M. Aiello and G. Ruffo, “LotusNet: Tunable privacy for distributed
OpenSocial of the state of neighborhood, (iii) tries to present a online social network services,” Computer Communications, Elsevier
B.V., vol. 35, no. 1, pp. 75-88. 2010.
consistent view, merging the data from the different sources
and (iv) it configures the discovering agents according to high [5] L.A. Cutillo, R. Molva and T. Strufe, “Safebook: a Privacy Preserving
Online Social Network Leveraging on Real-Life Trust,” in IEEE
level criteria, such as battery consumption and hardware Communications Magazine, vol 47, n.12, ser. Consumer
availability. The OpenSocial agent is basically a gateway to the Communications and Networking, 2009, pp 94-101.
OpenSocial container translating the other agents requests for [6] R. Baden, A. Bender, N. Spring, B. Bhattacharjee and D. Starin.,
the OpenSocial container and The TN agents is a true agent "Persona: an online social network with user-defined privacy", in
that performs potentially complex negotiations on his user’s Proceedings of the ACM conference on Data communication
behalf and depending from the configuration may work in (SIGCOMM '09), Barcellona, Spain: ACM, 2009, pp. 135-146.
entire autonomy. [7] M. Basuga, R. Belavic, A. Slipcevic, V. Podobnik, A. Petric and I.
Lovrek, “The MAgNet: Agent-based Middleware Enabling Social
Networking for Mobile Users,” in Proceedings of the 10th International
V. CONCLUSIONS Conference on Telecommunications ConTEL 2009 / Podnar Zarko,
Ivana; Vrdoljak, Boris, editor(s). Zagreb, Croatia: IEEE, 2009. 89-96.
In this paper we proposed a novel peer-to-peer social
networking platform that leverages existing, widespread and [8] F. Bellifemine, G. Caire, A. Poggi, G. Rimassa, “JADE: a Software
Framework for Developing Multi-Agent Applications. Lessons
stable technologies such as DHTs and BitTorrent. Although the Learned,” in Information and Software Technology Journal 50 (2008),
primitives offered by those technologies were created with pp. 10-21.
other goals in mind, however, they could be used with minor [9] D. Brickley and L. Miller, “FOAF vocabulary specification,”
modification in our system. In particular, we introduced a key- http://xmlns.com/foaf/0.1/, 2005.
based identity system and a model of social relations for [10] F. Bergenti, E. Franchi and A. Poggi, “Agent-based Social Networks for
distributing resources efficiently among interested readers. Enterprise Collaboration,” in Proceedings of the 20th IEEE
International Workshops on Enabling Technologies: Infrastructure for
In fact, we designed Blogracy as a micro-blogging social Collaborative Enterprises, 2011.
networking system, and we gave priority to the features more [11] A. Gursel and S. Sen, “Improving Search In Social Networks by Agent
important for micro-blogging, such as: (i) anonymity and Based Mining,” in IJCAI'09, Proceedings of the 21st international jont
resilience to censorship; (ii) authenticatable content; (iii) conference on Artifical intelligence, Morgan Kaufmann Publishers Inc.
semantic interoperability using activity streams and weak San Francisco, CA, USA, 2009, pp. 2034-2039.
semantic data formats for contacts and profiles; and (iv) data [12] B. Yu and M. P. Singh. “Searching social networks,” in AAMAS ’03:
availability. Proceedings of the second international joint conference on Autonomous
agents and multi-agent systems, New York, NY, USA, 2003. ACM Press,
After having implemented and tested all the core features of pp. 65–72.
Blogracy, we proved that a peer-to-peer architecture can be [13] F.E. Walter, S. Battiston and F. Schweitzer, A Model of a Trust-based
functional for both sharing files and advertising new social Recommendation System on a Social Network, Journal of Autonomous
activities. Moreover, by adhering to the OpenSocial standard, Agents and Multi-Agent Systems, vol. 16, no. 1 (2008), pp. 57-74.
the system can be integrated with other existing social [14] F. Bergenti, L. Rossi and M. Tomaiuolo, “Towards Automated Trust
platforms. Negotiation in MAS,” in Proceedings of the 10th Workshop “Dagli
Oggetti agli Agenti” (WOA 2009).
Apart from the core functions of the system, we are [15] L. Hamill and N. Gilbert, “Simulating large social networks in agent-
experimenting with more advanced features, which based models: A social circle model,” in Emergence Complexity
nevertheless are essential to provide a smooth experience to Organization (2010) Volume: 12, Issue: 4, Publisher: Emergent
users. In fact, differently from centralized systems, the nodes of Publications, pp. 78–94.
Blogracy are fully responsible for the platform operation. [16] F. Bergenti, E. Franchi and A. Poggi, “Selected Models for Agent-based
Friend discovery and content recommendation have to be Simulation of Social Networks,” in 3rd Symposium on Social Networks
realized in a completely distributed fashion, on the basis of and Multiagent Systems (SNAMAS '11), pp. 27–32.
trust agreements among users. [17] N. Li. “Local Names in SPKI/SDSI,” in Proceedings of the 13th IEEE
workshop on Computer Security Foundations (CSFW '00). Cambridge,
Finally, the realization of pervasive features on top of UK, July 2000: IEEE Computer Society Press, pp. 2–15.
Blogracy will require the dynamic management of open [18] J. Bethencourt, A. Sahai and B. Waters, “Ciphertext-Policy Attribute-
location and proximity groups, with possibly complex trust Based Encryption,” in IEEE Symposium on Security and Privacy, 2007.
negotiation protocols among autonomous agents. Oakland, California, USA, May 2007, pp.321-334.
[19] K. Rzadca, A. Datta and S. Buchegger, “Replica placement in P2P
storage: Complexity and game theoretic analyses,” in Proceedings of the
REFERENCES International Conference on Distributed Computing Systems. Genova,
[1] T. Berners-Lee, “Long Live the Web: A Call for Continued Open Italy: IEEE Computer Society, 2010, pp. 599–609.
Standards and Neutrality,” in Scientific American Magazine, December [20] J. Douceur and R. Wattenhofer, “Competitive hill-climbing strategies for
2010. replica placement in a distributed file system,” in Distributed
[2] S. Buchegger and A. Datta, “A case for P2P infrastructure for social Computing, ser. Lecture Notes in Computer Science, J. Welch, Ed.
networks – opportunities & challenges,” in Proceedings of Sixth Berlin / Heidelberg: Springer, 2001, vol. 2180, pp. 48–62.
International Conference on Wireless On-Demand Network Systems and [21] S. Bernard and F. Le Fessant, “Optimizing peer-to-peer backup using
Services (WONS 2009), Snowbird, UT, USA, feb. 2009, pp. 161–168. lifetime estimations,” in Proceedings of the 2009 EDBT/ICDT
[3] S. Buchegger, D. Schiöberg, L. Vu and A. Datta, “PeerSoN: P2P social Workshops. New York, NY, USA: ACM, 2009, pp. 26–33.
networking: early experiences and insights,” in Proceedings of the [22] E. Franchi, A. Poggi and M. Tomaiuolo, “Developing Applications with
Second ACM EuroSys Workshop on Social Network Systems. HDS,” in Proceedings of the 12th Workshop “Dagli Oggetti agli
Nuremberg, Germany: ACM, 2009, pp. 46–52. Agenti” (WOA 2011), pp. 117-122.