The Network is the People or The People are the Network (les gens sont le reseau) Jon.Crowcroft@cl.cam.ac.uk http://www.cl.cam.ac.uk/~jac22 Computer Laboratory University of Cambridge & Thomson Research Paris& LIP6 Dec 7, 2007 MSRC, Cambridge, UK Social Networks and Mobility are only just starting to be combined. • Two boring ways to do this are: – offering social network access on mobile device (merely a matter of coping with reduced network capacity and screen real estate, and somewhat different interaction paradigms); – making location services that are somewhat socially aware (what used to be called "contextual computing"). • Studies of human mobility patterns reveal that one can extract rich social structural information directly from contact (co-location) distributions. Dec 7, 2007 MSRC, Cambridge, UK This talk in outline • Thomson Labs: combine so-called traditional (legacy) social network structures (friends of friends on facebook or linkedin etc) with new mobility information (contact of contact in bluetooth or wifi encounters). • Two uses of the combined information: – expanding and subtly altering one's social net (allowing for the somewhat different nature of static references in online communities, from face-to-face encounters in Real Life); – using the social network (a priori, and learned, contacts and interests and tags) for efficient dissemination of information in wireless ad hoc communities. Dec 7, 2007 MSRC, Cambridge, UK Social Structures Vs Network Structures • Community structures – Social communities, i.e. affiliations – Topologically cohesive groups or modules • Centralities – Social hubs, celebrities and postmen – Betweenness, closeness, inference power, centrality • 1. Look at Structure of Human Social Mobility experimentally Dec 7, 2007 MSRC, Cambridge, UK Experimental setup • iMotes – ARM processor – Bluetooth radio – 64k flash memory • Bluetooth Inquiries – 5 seconds every 2 minutes – Log contact tuples: • {MAC address, start time, end time} Dec 7, 2007 MSRC, Cambridge, UK Experimental devices Dec 7, 2007 MSRC, Cambridge, UK So we did a lot of experiments • Mix of ours and other research groups: – – – – Infocom conferences 05 and 06 (miami, barcelona) In Cambridge (with 1st & 2nd yr undergrds) Hong Kong (with High School, and random [people on street) Plus others’ data • from crawdad dbase @dartmouth, ucsd, toronto • From MIT Reality Mining dataset • Typical data set involves: – 54 iMotes distributed, Experiment duration: 3 days – 41 yielded useful data – 11 with battery or packaging problem, 2 not returned Dec 7, 2007 MSRC, Cambridge, UK What we measure • For a given pairs of nodes: – contact times and inter-contact times. Duration of the experiment an inter-contact a contact time t Dec 7, 2007 MSRC, Cambridge, UK What we measure (cont’d) • Distribution per event. ≠ seen at a random instant in time. • Plot log-log distributions. • We aggregate the data of different pairs. (see the following slides). Dec 7, 2007 MSRC, Cambridge, UK Example: a typical pair α cutoff Dec 7, 2007 MSRC, Cambridge, UK Examples : Other pairs Dec 7, 2007 MSRC, Cambridge, UK K-clique Community Definition • • Union of k-cliques reachable through a series of adjacent k-cliques [Palla et al] Adjacent k-cliques share k-1 nodes Members in a community reachable through wellconnected well subsets Examples • • Overlapping feature Known Percolation threshold • • Dec 7, 2007 – 2-clique (connected components) – 3-clique (overlapping triangles) MSRC, Cambridge, UK K-clique Communities in Infocom06 Dataset Italian Paris Group A (French) Paris Group B (French) Barcelona Group (Spanish) K=5 Dec 7, 2007 MSRC, Cambridge, UK Visualisation – Evolution of Node Connectivity http://www.cl.cam.ac.uk/~ey204/Haggle/Vis/mobility.html Dec 7, 2007 MSRC, Cambridge, UK Other Community Detection Methodologies • Betweenness [Newman04] • Modularity [Newman06] • Information theory[Rosvall06] – Most literature is in social anthropology – Some in physics (AS level topology and web interconnect topology) – Future is to look at social net (myspace, linkedin, orkut, yahoo groups, facebook) – And user contributed tags (del.icio.us, etc) Dec 7, 2007 MSRC, Cambridge, UK Centrality in Temporal Network • Simulate flooding over the temporal graph • Uniform source/destination and temporal traffic distribution –Count number of times each node would be on shortest delay deliveries –Higher the count, more central node –Analogue to Freeman centrality • Need to threshold over lifespan of “link” in temporal graph Dec 7, 2007 MSRC, Cambridge, UK In-Group Centrality (Reality) Dec 7, 2007 Group A Group B Group C Group D MSRC, Cambridge, UK 2. And those online societies… • There’s a lot – you probably are on 1 – Hey, maybe you left several too • Anyone remember Usenet – Rec.*, Alt.* etc – The Well and Bboards Dec 7, 2007 MSRC, Cambridge, UK Lots of Social Networks • High Functionality – Facebook, Myspace – Content and applications shared as well as links and interests • Pure Discovery – Linkedin, Orkut – Mainly useful for finding work, Q&A etc Dec 7, 2007 MSRC, Cambridge, UK Lots of Real Social Groups • • • • • • Dec 7, 2007 Home, Family Work, Activities, Progress, Results, Admin Sport/Wellbeing Health, doctors etc Travel, Entertainment, shared interest Financial (pension schemes, saving, advice) MSRC, Cambridge, UK Internet Based Systems • Are large – Facebook cites 100s millions – Are connected – Serve some useful functions – Are differentiated (by functions, content) • Revenue can come from – advertisement – Or from spinoff activities – Or just bundled by ISP (remember AOL online and MiniTel Dec 7, 2007 MSRC, Cambridge, UK My LinkedIn Stats… Dec 7, 2007 MSRC, Cambridge, UK See Locale and Industry for me:- Dec 7, 2007 MSRC, Cambridge, UK Note two net characteristics • Very high node degree – 150 – is well known in sociology to be social group size • Very fast discovery – Within 1 month of being in Paris, already shows my network • Note this is on infrastructure/email/web based social net where I get about 1 join request per day – Mobile net would automate these joins – see next Dec 7, 2007 MSRC, Cambridge, UK Mobile Social Nets • Mobile social net tools/systems abound since recently • First generation • dodgeball – Nokia sensor, and others centered on cellular providers • Newer, trying to get into dating etc – – – – aka aki mobiluck meetmoi imity • Very new:- immediately useful (save phone state etc) – Mocospace – mig33 – zyb Dec 7, 2007 MSRC, Cambridge, UK But “sad” and geeky Dec 7, 2007 MSRC, Cambridge, UK What can we use to Hook? • Big Bug: – contradiction between social people meet people in “reallife” (RL) • and sad geeks, who inhabit virtual world:) • Possible escape routes: – – – – games (move from console to smart phone to console) music (listen and create) couch potatoe-able escape causes: view sport via multiple viewpoints and discuss • Other hooks (usefulness!): – disasters (data loss, infrastructure down) – cheap – s/w distribution Dec 7, 2007 MSRC, Cambridge, UK Building Mobile Network • Discover Common Locations (space) – Familiar Strangers • Fellow Commuters – Use location service • In network, or seperate • Discover Common Friends by Activity (time) – Friend of friend • By exchange of address books – Common interest • subscription or del.icio.us tags in common Dec 7, 2007 MSRC, Cambridge, UK Content Distribution on Mobile Network • Use infrastructure • GPRS/Edge/3G – Or WiFi Hotspots • Ad Hoc Networking – MANET forward via others smart phones • Delay Tolerant Networking – Store, Carry, Forward • In all these, need to know where humans are and when Dec 7, 2007 MSRC, Cambridge, UK The people are the network • The mobility of people is input to network design – Either to place capacity for infrastructure, – Or for setting up regular, or opportunistic forwarding paths over users devices – Need to distunguish between people who are social butterflies (hub for idle chit chat/gossip) • Meet lots of people – and people who are central for forwarding a lot of information • Meet people and exchange a lot of data for them often Dec 7, 2007 MSRC, Cambridge, UK Problems • Fear • Users fear: – spam, loss of euros, loss of battery life, identity theft, denial of service, • Too many interrupts to life – Too little privacy • Does community (e.g. shared interest or locale) help? … Dec 7, 2007 MSRC, Cambridge, UK What do we Trust • Users trust members of their tribe a lot more than strangers, but… • To build a network, usually we want bidirectional “links” • Are trust relationships associative, commutative, transitive, reflexive? • Not in social relationships, usually: Dec 7, 2007 MSRC, Cambridge, UK Can we build incentives? • Perhaps additional capacity (mutual benefit) is enough? – potential capacity for DTN use of multihop, store-carry-forward much more than cellular – But delay highly unpredictable and possibly hours or days (although we tolerate this for email) – Can we enhance delay and trust with infrastructure? Dec 7, 2007 MSRC, Cambridge, UK Use infrastructure to bootstrap • Yes, infrastructure helps a lot – Can bootstrap social network from infrastructure – Can bootstrap payment and identity (SIM card and cellular or WiFi contract) from infrastructure – Can use nodes with infrastructure access to shortcut, and improve delay distributions • Trade off against a cost Dec 7, 2007 MSRC, Cambridge, UK Could even monetize call time • E.g. I swap carrying mp3s for you, for you giving me minutes of voice call time on (say) SFR or Vodafone • I can ask infrastructure to help with – Location service – Channel/spectrum allocation for MANET/DTN Users – Proof of Identity – Authenticating “currency” (a “mint”) Dec 7, 2007 MSRC, Cambridge, UK Thank you • Merci pour votre attention • Questions? • Jon.Crowcroft@cl.cam.ac.uk • Next slides show that we can build social networks in a decentralized way • Without any infrastructure – • future work would include hybrid with infrastructure (to boot, and to provide persistence) and garbage collection Dec 7, 2007 MSRC, Cambridge, UK More Future Work… • Privacy and trust (including incentives forwarding). • Absolute location based, versus co-location based forwarding • Other layer social network based forwarding, e.g • Content (url distance, or meta, e.g. del.icio.us tag) based forwarding • Myspace, Orkut, etc Dec 7, 2007 MSRC, Cambridge, UK Jon.Crowcroft@cl.cam.ac.uk et merci a CNRS et Thomson Dec 7, 2007 MSRC, Cambridge, UK