Presentation is loading. Please wait.

Presentation is loading. Please wait.

Stimulating Participation in Wireless Community Networks Elias C. Efstathiou, Pantelis A. Frangoudis, George C. Polyzos Mobile Multimedia.

Similar presentations


Presentation on theme: "Stimulating Participation in Wireless Community Networks Elias C. Efstathiou, Pantelis A. Frangoudis, George C. Polyzos Mobile Multimedia."— Presentation transcript:

1 efstath@aueb.gr Stimulating Participation in Wireless Community Networks Elias C. Efstathiou, Pantelis A. Frangoudis, George C. Polyzos Mobile Multimedia Laboratory Department of Computer Science Athens University of Economics and Business IEEE INFOCOM 2006, Barcelona, April 23-29, 2006

2 efstath@aueb.gr  Connected to DSL and cable links  Secured (WEP, WPA)  Under-utilized Motivation Many private WLANs in cities An opportunity?  E.g. Skyhook Wireless WPS is a positioning service that relies on detecting catalogued WLANs Boston WLAN coverage © 2005 Skyhook Wireless Inc Manhattan WLANs © 2002 Public Internet Project.org

3 efstath@aueb.gr Outline of Talk 1)The Peer-to-Peer Wireless Network Confederation scheme for WLAN sharing 2)P2PWNC architecture and algorithms 3) Simulation-based evaluation of algorithms 4) P2PWNC protocol 5) P2PWNC implementation

4 efstath@aueb.gr What is P2PWNC ? 3)A public good created through private contributions  Wireless Internet (and VoIP) through third parties  And what about freeriders? 2)A simple protocol for (selfish) WLAN phones and access points  Our implementation runs on Windows Mobile phones/PDAs and Linux access points 1)An incentive scheme for sharing our WLANs with others  Relies on direct and/or cyclical reciprocity  Balances consumption and contribution AB A C B

5 efstath@aueb.gr Distinctive Characteristics 1)Complete decentralization, respect for privacy  Open to all, without central authorities of any kind  Free identities (aka cheap pseudonyms)  No tamperproof modules 2)Selfishness  Nobody cares about the good of the community  Freeriding is rational, and it extends to accounting tasks  No super peers, no P2P overlaysP2PWNC ?sophisticated freeridersAttacker model centralizeddecentralizedArchitecture WLAN sharing Goal versus

6 efstath@aueb.gr Caveats 1)ISP prohibitions on connection sharing 2)People living in the outskirts… what about them? 3) Designed for citywide everyday WLAN roaming, as opposed to global WLAN roaming 4) WLAN standards do not support fast handovers, yet 5)We do not defend against pure irrational malice (i.e. DoS attacks at various layers)

7 efstath@aueb.gr System Elements 1)Identities  System identities are public-private key pairs  Standard cryptographic assumptions apply 3) Receipts  Generated each time WLAN service is contributed and consumed  Digitally signed by the consumer consumer public key timestamp signed using consumer private key weight provider public key 2)P2PWNC peers  Each with a self-generated identity  At least one WLAN access point per peer (for contributing to the community)  At least one WLAN client per peer (for consuming from the community)

8 efstath@aueb.gr Receipt Graph w 14 A B C G H F E D I w1w1 w2w2 w3w3 w4w4 w5w5 w6w6 w7w7 w8w8 w9w9 w 10 w 11 w 12 w 13 1)A (logical) directed graph, with cycles  Vertices represent peer IDs  Edges point from the consumer to the provider (they represent debt)  Edge weight is the sum of the weights of the corresponding receipts 2)Graph security  Freeriders and colluders can create an arbitrary number of vertices and edges (Sybil/false trading attack)  They cannot create fake edges starting from peers who are not colluding 3)Receipt repository  Centralized version  Distributed version  Short-term history, in both versions

9 efstath@aueb.gr Reciprocity Algorithm: Overview mf(P→C) maxflow from P (Provider) to C (Consumer) gmf(P→C) generalized maxflow from P to C: discount sub-flows, multiply them by 2 (1—path_length) quality of flow from P to C qf avg an average qf for the community, estimated independently by each peer P, using: qf avg ← a × qf avg + (1–a) × qf(P→C), 0 < a < 1 r = r 1 × r 2 benefit function

10 efstath@aueb.gr Gossiping and Bootstrap At one extreme, APs would only store graph edges pointing directly to them At the other extreme, a central server would store everything 1)How do we circulate receipts through the system?  Gossip a)Copy “fresh” receipts from your WLAN AP to your phone b)Present them to foreign APs before requesting service c)Foreign APs merge these with their local receipt graph …verification overhead, time horizon, incentive compatibility 2)How to bootstrap a peer?  Reciprocity Algorithm does not work without unconditional cooperation in the beginning  Define patience: the number of successful consumptions after which a new peer starts to use the Reciprocity Algorithm

11 efstath@aueb.gr Evaluation Framework Evolutionary simulations  Uniform peers (consumption rate, benefit and cost functions, repository sizes, merging sizes, patience)  Assume unit weight on receipts, corresponding to a “standard WLAN session”  Constant provider cost (per session), normalized to c = 1  Variable consumer benefit (b min = 0, r th = 0.1, b max = 6)  Rounds: symmetric matches  Strategy ratings  Assume rational but myopic peers: “trial and error”  Strategic mixtures, and learning and mutation  Goal: the evolution of cooperation (average rating near b max – c)  No congestion effects  Also simulate system growth (up to n peers)

12 efstath@aueb.gr 100 200 300 400 500 Rounds Merging and Patience Community grows to n = 100 peers (at round 100) and private receipt repositories contain 1000 receipts each diminishing returns patience=10 patience=100

13 efstath@aueb.gr 100 200 300 400 500 600 700 800 900 1000 Rounds Evolutionary Learning Community grows to n = 100 peers (by round 100) and private receipt repositories contain 1000 receipts each Non-zero learning and mutation probabilities; merge 100 conditions favor freeriding mutations set b max /c = 3 (was: 6) P2PWNC service not valuable enough, ALLD wins

14 efstath@aueb.gr Change ID to Erase Debt (i.e. r = r 1 ) r = r 1 × r 2 qf avg ← a × qf avg + (1–a) × qf(P→C), 0 < a < 1 B C CiCi C i+1 A P hiding in qf avg ’s variance, but still, worse off Community starts with n = 200 peers and private receipt repositories contain 2000 receipts each; merge 200

15 efstath@aueb.gr Simple text-based protocol over TCP/IP The P2PWNC Protocol C P CONN CACK 11:50am = t 0 (consumer connects) C P RREQ RCPT 11:51am (P requests 1 st receipt) RCPT timestamp = t 0 RCPT weight = w 1 C P RREQ RCPT 11:52am (P requests 2 nd receipt) RCPT timestamp = t 0 RCPT weight = w 2 > w 1 P RREQ RCPT 11:53am (consumer has departed) P stores last receipt (timeout) Receipt Repository

16 efstath@aueb.gr P2PWNC Implementation  Open source: downloadable from http://mm.aueb.gr/research/P2PWNC  Linksys-centric all-in-one version  ECC and RSA digital signatures – tradeoffs  Guru-friendly modular version in preparation, for deployment in AWMN  P2PWNC client for Windows Mobile WLAN phones  Plus, VPN tunnels (PPTP and L2TP/IPsec), QoS module, VoIP proof-of-concept app.

17 efstath@aueb.gr Thanks! Elias C. Efstathiou Mobile Multimedia Laboratory Department of Computer Science Athens University of Economics and Business efstath@aueb.gr http://mm.aueb.gr/research/P2PWNC


Download ppt "Stimulating Participation in Wireless Community Networks Elias C. Efstathiou, Pantelis A. Frangoudis, George C. Polyzos Mobile Multimedia."

Similar presentations


Ads by Google