Are You moved by Your Social Network Application? Abderrahmen Mtibaa, Augustin Chaintreau, Jason LeBrun, Earl Oliver, Anna-Kaisa Pietilainen, Christophe Diot Thomson Paris Research Lab Avinash Patlolla

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Introduction Before the Internet: socialize by physical meeting Today: Internet allows “virtual” socializing Tomorrow: Meet the virtual community using opportunistic contacts and locality

Motivation Explore the relation between virtual social interactions and human physical meetings Understand complex temporal properties based on simple social properties Forwarding based on social network properties

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Experimental Setup Distributed smartphones with mobile opportunistic social networking application to 28 participants (but later reduced to 27) 3 days experiment at CoNext December ‘07 Initially, each participant was asked to identify friends among the 150 CoNext participants. Applications: ◦ Opportunistic socializing: make new friends based on friends ◦ Asynchronous messaging

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Terminology and Definitions Social graph: Graph of friendship between participants. Denoted as G = (V, E) Contact graph: Collection of opportunistic Bluetooth contacts between the participants form the temporal network, which is called contact graph. Denoted as G t = (V, E t ) Delay- optimal path: A path, in contact graph, from s to d starting at time t0 is delay- optimal if it reaches the destination d in the earliest possible time. The delay optimal path can be computed via dynamic programming

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Topological comparison Initial GraphFinal Graph # nodes # edges Average degree Diameter

Node properties Characterize Node heterogeneity ◦ High/low activity ◦ Popularity ◦ Contact rate Two metrics are measured ◦ Node degree  Social graph: number of friends  Contact graph: average number of devices seen per scan (every 2 min.) ◦ Centrality of nodes:  Social graph: measure the occurrence of the node inside all shortest paths  Contact graph: measure the occurrence of the node at each time t inside all shortest paths

Node degree Ordering error =

Centrality of nodes

Contact properties Compare contact according to: ◦ Social distance (friends have distance 1, friends of friends of friends have distance 2 and so on) ◦ time between two successive contacts

Path properties Delay-optimal paths as a function of the social distance between the source and the destination

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Social forwarding paths General model: depending on the source s and the destination d, a rule defines a subset of directed pairs of nodes (u  v) so that only the contacts occurring for pairs in the subset are allowed in forwarding path. Path construction rules: ◦ neighbor(k): (u  v) is allowed if and only if u and v are within distance k in the social graph ◦ destination-neighbor(k): (u  v) is allowed if and only if v is within distance k of d ◦ non-decreasing-centrality: (u  v) is allowed if and only if C(u) <= C(v) ◦ non-increasing-distance: (u  v) is allowed if and only if the social distance from v to d is no more than the one from u to d

Performance of different path construction rule

Comparison of rules The neighbor rule performs reasonably well The rule based on centrality outperforms all the other rules considered The combination of neighbor and centrality rules reduce the cost (offers best trade-off)

Outline Introduction ◦ Motivation Experimental setup Terminology and Definitions Topological comparisons ◦ Node properties ◦ Contact properties ◦ Path properties Social forwarding paths Summary and critiques

Summary and Critiques Similarity in the properties of nodes, contacts and paths in the two graphs Highlighted the importance of centrality of nodes Using social neighbors to communicate can be effective to exchange messages with opportunistic bandwidth Critiques: ◦ Important issues not yet studied, like computing centrality of nodes in a distributed manner. ◦ Scalability and usability ◦ Not many technical details

References Questions???