A survey on the state-of-the-art of social communication patterns and opportunistic forwarding Emmanouil Dimogerontakis, Antonio Severien and Faik Aras Tarhan @RALIAS
May 11, 2015
A survey on the state-of-the-art of social communication
patterns and opportunistic forwarding
Emmanouil Dimogerontakis, Antonio Severien and Faik Aras Tarhan
@RALIAS
Outline
● Intro● Useful Knowledge● Social Patterns and Opportunistic
Forwarding● Evaluation of state of the art● Conclusions
Intro
Opportunistic networking targets infrastructureless environments where mobile nodes wish to communicate with each other in
highly dynamic and unpredictable topology
Intro
Knowledge of social behaviour can be used to enhance and fine tune performance on
opportunistic mobile networks
Challenges
● Increased mobility of nodes
● Mobility traces that combine traffic and social information are rare
● Artificially generated simulation environments are not a good replacement for real world scenarios
● Message delivery in DTNs can vary from minutes to days
● Influence of the nature of human interactions
Main Approaches
Three main questions: ● Which is human social behaviour under a given
circumstance? ● Does social behaviour affect the network performance? ● How can we exploit existing social and mobility
information (social graph, contact history) to enhance the network performance and resource usage?
Outline
● Intro● Useful Knowledge● Social Patterns and Opportunistic
Forwarding● Evaluation of state of the art● Conclusions
Social Behaviour
Community: indicate one’s social roleCentrality: reflects authority or popularity in a group
○ Degree: degree of an individual node ○ Closeness: social distance○ Betweenness: the relay capability, or
“interpersonal influence”, of nodesTie-strength: the robustness of relationship for a dyad
○ Frequency○ Recency○ Duration
Similarity: strength of a common attribute(Similar communication patterns)
Social Network Models
Ways to construct graphs with communities.Caveman Model:
○ initially K fully connected graphs, then every edge of the initial network is re-wired to point to a node of another cave with a certain probability p
○ able to reproduce social structures very close to real ones.
Kumpula Model:○ the weights are generated dynamically and shape
the developing topology○ local attachment, global attachment
Outline
● Intro● Useful Knowledge● Social Patterns and Opportunistic
Forwarding● Evaluation of state of the art● Conclusions
Social Patterns and Opportunistic Forwarding
Which is human social behaviour under a given circumstance?
Stumbl: Using Facebook to collect rich datasets for opportunistic networking research
Deals with understanding the fundamental patterns of human mobility, social relations and communications in order to create algorithms and protocols that exploit human mobility and consequent wireless contacts for better dissemination.
● Unlike “Mobiclique: Middleware for mobile social networking,” handling only one or two of the aspects of relations, this paper focuses on all three combined
Stumbl: Using Facebook to collect rich datasets for opportunistic networking research
Results:
● Social tie type has very strong impact on meeting characteristics in terms of context, duration and frequency of meetings
● The type of social tie has strong impact on context, duration and frequency of meetings
● The number of Facebook communication events differs for different relationship ties
● People communicate preferentially with friends they also have face- to-face meetings.Thus, communication ties are more local than social ties
Stumbl: Using Facebook to collect rich datasets for opportunistic networking research
Criticism:● Meetings and communication parts are vulnerable to
Stumbl users’ misleading information since it is self-reporting
● Running bigger Stumbl experiments with more participants should be the next step
● Needs to provide incentives to the users to regularly report true data about their face-to-face meetings
● Creating a more efficient algorithm or protocol for opportunistic networking
Social Patterns and Opportunistic Forwarding
Does social behaviour affect the network performance?
Dissemination in Opportunistic Mobile Ad-hoc Networks: the Power of the Crowd
Studies fundamental properties of human interactions. Nodes not showing up in the network frequently or periodically might play the major role in data dissemination depending on the characteristic of the network
● “Bin” method observing whether people’s mobility patterns exhibit a diurnal behavior to:○ Classify the users as Vagabonds or Socials
Simulation:
● Metrics: Contamination● Mobility Traces: The Dartmouth data set, The San
Francisco data set, The Second Life data set
Dissemination in Opportunistic Mobile Ad-hoc Networks: the Power of the Crowd
Results:
● Vagabonds eventually dominates dissemination using Socials if and only if
● The effectiveness of contamination is more a matter of contact “density” in an area than an issue of social behavior
● Vagabonds have an important role in dissemination of information and should not be ignored unlike papers tending to neglect this kind of users such as:○ “PeopleRank: Social Opportunistic Forwarding”○ “Social-Based Trust in Mobile Opportunistic
Networks”
Dissemination in Opportunistic Mobile Ad-hoc Networks: the Power of the Crowd
Criticism:● They merely focus on flooding routing:
● Message transfers are assumed to be instantaneous● Assumption that contacts take place between any two
devices associated to the same access point is not enough to represent the reality in fact
● Investigating the interactions between Vagabonds and Socials in supporting information dissemination
● Investigating the dynamics of user social behavior with respect to different social communities as done in paper “SREP routing in opportunistic network”
The effect of communication pattern on opportunistic mobile networks
How social communication patterns which are based on basic metrics of theory of sociology affect the behaviour of the opportunistic mobile networks.Social patterns:● Community-biased● Centrality-biased (degree, closeness, betweenness)● Tie-strength-biasedRouting algorithms with social utilities:● Prophet (contact frequency)● SimBet (betweenness centrality, similarity)● FairRouting (aggregated interaction strength)
The effect of communication pattern on opportunistic mobile networks
Simulation:● Metrics: Success rate● Mobility Traces: Reality Mining (MIT) and Haggle
(Infocom 2006)● Community and Social information for datasets:
Constructed with community detection tool CFinderResults:● Social-based communication patterns increase the
system throughput of social-based routing protocols● Tie-strength-biased offers the best performance● Network topology can greatly influence network
performance (centrality-biased, community-biased)
Social-Based Trust in Mobile Opportunistic Networks
A real-trace driven approach to study the tradeoff between trust and success delivery rates in opportunistic networks. Potential impact of excluding a few popular nodes from the opportunistic forwarding can be solved by enabling trust across communicating entities and integrating incentives into the operation of opportunistic networks.Social-Based Trust Filters:● Relay-to-Relay, Source-to-Relay● Social Estimators: -d-distance (d is a parameter)
-Common interests -Common Friends -Combination
Social-Based Trust in Mobile Opportunistic NetworksSimulation:● Metrics: normalized success rate within time t,
normalized cost (i.e. # of replicas)● Mobility Traces: CoNext07, CoNext08, Infocom06● Community and Social information for datasets:
available from the experiment or obtained offlineResults:● S2R filters success rate increases linearly with the cost● R2R filters achieve better performance than S2R,
which is performing poorly● Best R2R filter: combination 1-distance and common
friends● The common friends technique appears to be the best
from the ones proposed
Selfishness, Altruism and Message Spreading in Mobile Social Networks
Evaluate using real traces how robust an opportunistic network is under different distributions of altruism in the population. Social patterns:● Altruism Distributions: percentage of selfishness,
uniform, normal, geometric, degree-biased, community-biased
Communication patterns:● Uniform (evaluate with datasets)● Community-Biased (evaluate with static social network
models)
Selfishness, Altruism and Message Spreading in Mobile Social Networks
Static Social network models:● Caveman model● Kumpula modelSimulation:● Metrics: delivery/success ratio● Mobility Traces: Reality Mining (MIT) Cambridge,
Infocom05, Infocom06● Simulator: Contact-driven● Community and Social information for datasets: not
complete
Selfishness, Altruism and Message Spreading in Mobile Social Networks
Results:● Opportunistic networks generally be robust against
altruism● Main cause of robustness: multiple forwarding paths● Traffic pattern chosen for simulation has significant
impact on the social behavior impact of the simulated network
Social Patterns and Opportunistic Forwarding
How can we exploit existing social and mobility information (social graph, contact
history) to enhance the network performance and resource usage?
PeopleRank: Social Opportunistic Forwarding
Like a distributed PageRank, PeopleRank identifies the most popular nodes (in a social context) to forward the message to, given that popular nodes are more likely to meet other nodes in the networks.Social patterns:● People/nodes are ranked as “important” when they
are linked in a social context to many other “important” people
● Centralized and distributed versionRouting algorithm:● A node u forwards data to a node v that it meets if the
rank of v is higher than the rank of u.
PeopleRank: Social Opportunistic Forwarding
Simulation:● Metrics: average message delivery delay, overhead or
cost by mechanism (i.e. # of replicas)● Mobility Traces: MobiClique, SecondLife, Infocom06
(interest,facebook,union), and Hope● Community and Social information for datasets: some
explicit, some implicit
PeopleRank: Social Opportunistic Forwarding
Results:● forward to socially best nodes improves overall success
rate● outperforms simple social forwarding algorithms and
some of the well-known contact-based algorithms (i.e. Spray & Wait)
● End-to-end delay and a success rate close to those given by flooding while reducing the number of retransmission by 50%
Social relationship enhanced predictable routing in opportunistic network
Network is composed of communities and nodes are assumed to roam among communities somewhat regularly. To introduce this mobility of the node, semi-deterministic Markov process modelling is adapted and to quantify the social degree of the node, PageRank algorithm is introduced.
● PageRank algorithm is adapted to evaluate social ranking of the nodes in the same community to calculate the centrality of the nodes
● Every node in the same community has a unique social degree
Social relationship enhanced predictable routing in opportunistic network
● the total prediction correction of social degree of a node with all communities at time t
● the average prediction correction of social degree of node
Simulation:Metrics: Delivery Delay, Delivery Ratio, Time To Live (TTL), Deviation Degree● There are several predefined communities in the
network.● Visits are probabilistic and self-determined.
Simulator: ONE
Social relationship enhanced predictable routing in opportunistic network
Results:
● The efficiency of SREP algorithms is acceptable, when the randomness of the node deviation is lower.
● When the TTL is longer enough, the performance of every routing improve
● SREP makes full use of the feature of human society, and coincides the mobility of the human mobility
● SREP can yield the improvement of the delivery ratio and reduce the delivery delay in some defined scenario
Forming a Social Structure in Mobile Opportunistic NetworksThey exploit the mobile nodes frequency interactions to form social structures in opportunistic networks by understanding the relationship between the mobile nodes.Methods:
○ Social Structure based on Average Frequency Interactions■ measures how many times the same pair of nodes are
co-located and interact within a given period of time○ Social Structure based on Periodicity Frequency
Interactions■ based on the interactions frequency that occur in a
given period of time○ Social Structure based on Sliding Window
■ Sliding Window (SW) is a frame that subdivided into number of slots, which is a single time step in period
Forming a Social Structure in Mobile Opportunistic Networks
Simulation:
● Metrics: In Degree and Out Degree links, Threshold ● Simulator: UCINET Criticism:
● Mobility in the simulation is based on Random Walk. It does take human social contact incentives into account ○ unlike paper “Social relationship enhanced
predictable routing in opportunistic network”.
Forming a Social Structure in Mobile Opportunistic Networks
Results:
● The formation of social structure is depended on the policy of the node interactions
● A social structure of nodes is different at different point of time
● Social Structure based on Sliding Window, is more appropriate to be deployed as the formation of the social structures are dynamic and represent the current nodes interaction in which represent the underlying current network topology
Bootstrapping Opportunistic Networks Using Social Roles
Proposes Social Role Routing (SRR)Bootstrap an opportunistic network without node contact information from Self-Reported Social Networks (SRSN)Avoid overloading popular nodes Social Patterns:
● Define roles for nodes where nodes communicate in same social classes
Routing Algorithms with social utilities:
● Social Role Routing (SRR) takes advantage of roles grouping to make forwarding decisions
Bootstrapping Opportunistic Networks Using Social Roles
Sending messages from group A to B - Node 1 might be overloaded, use node 12 and 13
Bootstrapping Opportunistic Networks Using Social Roles
Routing protocol evaluation
● Epidemic: forward to any encountered node● SimbetTS: contact history based (warm-up time)● Social Role Routing (SRR): forward message to similar
roles● Social Role Routing SimbetTS Hybrid: switches from
SRR to SimbetTS
Optimizing Message Delivery in Mobile Opportunistic Network
Nile routing protocolUse of replicas to increase delivery probabilityCompromise between flooding and intelligent routing techniques
- Replicate aggressively in sparse networks- Restrict replication on dense networks- Considers congestion control to determine replication
Social Patterns:
● Routing is flexible to adapt to different social patternsRouting Algorithms with social utilities:
● Utilises contact frequency
MobiClique: Middleware for Mobile Social Networking
Mobile social software to maintain and extend online social networks through opportunistic encounters in real-lifeMiddleware to build apps on top
- Neighborhood discovery- User identification- Data exchange
Social Patterns:
● Monitors mobility and social behaviorRouting Algorithms with social utilities:
● Opportunistic forwarding
CAMEO: Context-Aware Middleware for Opportunistic Mobile Social Networks
Management, elaboration and dissemination of context informationIdentification of context components through hash values Social Patterns:
● Social context Routing Algorithms with social utilities:
● Publish/Subscribe between interest groups● Beaconing mechanism to find relevant context● Evaluates the probability of each neighbor node to
deliver the message to destination
Outline
● Intro● Useful Knowledge● Social Patterns and Opportunistic
Forwarding● Evaluation of state of the art● Conclusions
Evaluation
● Similar data traces - there is a need for more experimentation
● Similar references - base knowledge from same sources
● Contradiction between papers. For example:
- [9],[4]: focus on unpopular nodes importance- [6],[8]: focus on enhancing popular nodes
● Improvements. For example:
- [12] adds community idea in [6] with social rank
Outline
● Intro● Useful Knowledge● Social Patterns and Opportunistic
Forwarding● Evaluation of state of the art● Conclusions
Conclusions
Social awareContext awareMobility awareNetwork aware
Improvement of opportunistic forwarding protocols
Future
● Power consumption related to social behavior
● Devices are now ad hoc compatible (WiFi)● Marketing oriented social behavior on
MANETS● A lot of ongoing research (SOCIALNETs etc.)
References1. Islam, M.A.; Waldvogel, M.; , "Optimizing message delivery in mobile-opportunistic networks," Internet Communications (BCFIC Riga), 2011 Baltic Congress on Future , vol., no., pp.134-141, 16-18 Feb. 20112. Anna-Kaisa Pietilinen, Earl Oliver, Jason LeBrun, George Varghese, and Christophe Diot. 2009. MobiClique: middleware for mobile social networking. In Proceedings of the 2nd ACM workshop on Online social networks (WOSN '09). ACM, New York, NY, USA, 49-54.3. Arnaboldi, V.; Conti, M.; Delmastro, F.; , "Implementation of CAMEO: A context-aware middleware for Opportunistic Mobile Social Networks," World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2011 IEEE International Symposium on a , vol., no., pp.1-3, 20-24 June 20114. Bigwood, G.; Henderson, T.; , "Bootstrapping opportunistic networks using social roles," World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2011 IEEE International Symposium on a , vol., no., pp.1-6, 20-24 June 20115. Xiaoguang Fan; Kuang Xu; Li, V.O.K.; Guang-Hua Yang; , "The effect of communication pattern on opportunistic mobile networks," Consumer Communications and Networking Conference (CCNC), 2011 IEEE , vol., no., pp.1016-1020, 9-12 Jan. 20116. Mtibaa, A.; May, M.; Diot, C.; Ammar, M.; , "PeopleRank: Social Opportunistic Forwarding," INFOCOM, 2010 Proceedings IEEE , vol., no., pp.1-5, 14-19 March 20107. Pan Hui; Kuang Xu; Li, V.O.K.; Crowcroft, J.; Latora, V.; Lio, P.; , "Selfishness, Altruism and Message Spreading in Mobile Social Networks," INFOCOM Workshops 2009, IEEE , vol., no., pp.1-6, 19-25 April 20098. Mtibaa, A.; Harras, K.A.; , "Social-Based Trust in Mobile Opportunistic Networks," Computer Communications and Networks (ICCCN), 2011 Proceedings of 20th International Conference on , vol., no., pp.1-6, July 31 2011-Aug. 4 20119. Zyba, G.; Voelker, G.M.; Ioannidis, S.; Diot, C.; , "Dissemination in opportunistic mobile ad-hoc networks: The power of the crowd," INFOCOM, 2011 Proceedings IEEE , vol., no., pp.1179-1187, 10-15 April 201110. Lenando, H.; Zen, K.; Jambli, M.N.; Thangaveloo, R.; , "Forming a Social structure in mobile opportunistic networks," Communications (APCC), 2011 17th Asia-Pacific Conference on , vol., no., pp.450-455, 2-5 Oct. 2011
References11. Hossmann, T.; Legendre, F.; Nomikos, G.; Spyropoulos, T.; , "Stumbl: Using Facebook to collect rich datasets for opportunistic networking research," World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2011 IEEE International Symposium on a , vol., no., pp.1-6, 20-24 June 201112. Xie, X., Zhang, Y., Dai, C., & Song, M. (2011). Social Relationship Enhanced Predicable Routing in Opportunistic Network. 2011 Seventh International Conference on Mobile Ad-hoc and Sensor Networks, 268-275. 13. http://www.haggleproject.org/14. http://reality.media.mit.edu/15. http://www.social-nets.eu/16. http://crawdad.cs.dartmouth.edu/17. S. Wasserman and K. Faust, Social network analysis: methods and applications, Cambridge University Press, 199418. J. M. Kumpula, J. P. Onnela, J. Saramaki, K. Kaski, and J. Kertesz. Emergence of communities in weighted networks. 200719. D. J. Watts. Small Worlds The Dynamics of Networks between Order and Randomness. Princeton Studies on Complexity. Princeton University Press, 1999
A survey on the state-of-the-art of social communication
patterns and opportunistic forwarding
Emmanouil Dimogerontakis, Antonio Severien and Faik Aras Tarhan
@RALIAS