1. Elias C. Efstathiou Mobile Multimedia Lab Computer Science Department Athens University of Economics and Business Athens 10434, Greece efstath@aueb.gr - http://mm.aueb.gr Authors: P. Antoniadis, C. Courcoubetis, E. C. Efstathiou, G. C. Polyzos, and B. Strulo IST Project MMAPPS - Market Management of Peer-to-Peer Services (RTD No IST-2001-34201) IST Mobile and Wireless Summit 2003, Aveiro, Portugal The Case for Peer-to-Peer Wireless LAN Consortia (PWC) A P2P Approach to WLAN Roaming

2. 2 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr Introduction (1 of 2) Ubiquitous Internet Access a Necessity However, WISPs are facing difficulties WISP roaming practically non-existent Many under-exploited private WLANs do exist The Peer-to-Peer Wireless LAN Consortium (PWC): A Framework for uniting all WLANs in one global group A Community of WLAN Administrative Domains that offer wireless Internet access to each other’s registered users The PWC is a P2P network of Domain Agents (DAs) DAs are physical nodes that represent one domain each Their purpose is to eliminate the overhead of roaming agreements Instead, DAs obey a simple token-exchange rule

3. 3 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr Introduction (2 of 2) Domain Independence: a PWC Distinctive Characteristic DAs make autonomous decisions concerning the amount of resources they provide to visitors Key difference from other roaming schemes PWC Simplicity No central entity controls the PWC or the interactions of its participants No cost of entry for domains PWC subsystems leverage its P2P nature: no external servers are required

4. 4 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr Background Motivation Existing under-exploited WLANs IEEE 802.11 simplicity Next-generation portable devices WLAN Roaming Today Practically non-existent Hotspot aggregation (e.g. Boingo Inc.) is not WLAN roaming Limitations of WISP associations (e.g. Pass-One) Service-mark logic Insufficient privacy Insufficient autonomy Administrative overhead and complexity The PWC as a P2P System Shared good: bandwidth Autonomous peers: independent domain agents Free-riding: domains that may not provide access to visitors Incentives and rules: token-exchange rule

5. 5 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC Requirements 1. Domain Independence The peers make autonomous decisions Concerning their contribution level Concerning their participation status 2. Domain Reciprocal Behavior Free-riding must be minimized PWC system rule: token-exchange This rule “guides” domain behavior 3. Easy-to-Join No administrative overhead Similar to joining a P2P file-sharing network Assuming the domain WLAN infrastructure is already in place 4. PWC Self-Sufficiency PWC subsystems rely only on the PWC peers themselves 5. Decentralization No central entities No central authority manages the PWC

6. 6 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC Entities Domain Agents (DAs) The PWC is a P2P network of DAs DAs are nodes running the PWC DA software Exactly one DA per PWC administrative domain Each DA has a unique logical name: aueb.gr, cometa.net The_Aveiro_Smith_Family Users Registered with one (or more) DAs Each has a unique identifier ( user_name@logical_domain_name ) Bandwidth The PWC ‘good’ - 802.11 bandwidth and bandwidth to the Internet Tokens Unforgeable virtual currency Exchanged between DAs Represent the value, which DAs ascribe to their consumed bandwidth

7. 7 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC High-Level View WLAN view P2P view AP : WLAN Access Point : User DA : PWC Domain Agent APAP APAP APAP APAP APAP APAP APAP APAP APAP DA ‘White’ DA ‘Gray’ DA ‘Black’

8. 8 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC Domain Agent Modules 1. Name-service Maps logical domain names to DA IP addresses Uses a Distributed Hash Table (DHT) 2. Authentication Maintains a database of registered users Along with their security credentials 3. Traffic Policing Logs and shapes egress and ingress Internet traffic Allocates specific amounts of bandwidth to visitors 4. WLAN Firewall, DHCP, DNS, NAT/NAPT, WLAN control 5. Distributed Accounting Secure storage of PWC accounting information Also uses a DHT 6. Consumer Strategy Regulates the consumption actions of the domain’s roaming users 7. Provider Strategy Regulates contribution to visitors Dynamically assigns “prices” to consumed resources 8. Privacy Enhancement Ensures PWC user anonymity and untraceability

9. 9 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC Security Issues PWC security is a superset of WLAN security The usual confidentiality, integrity, and availability problems apply The following three issues are PWC-specific: 1. Traffic Logging by Untrustworthy Providers User traffic completely visible to the visited Domain Agent Encryption does not hide useful metadata (e.g. remote-party address) SOLUTION: Tunnel (encrypt and route) through the home DA 2. Identity Privacy: PWC Pseudonyms User name visible to the visited DA SOLUTION: Use algorithmically updated user aliases 3. Anonymity and Untraceability: PWC Mixes User name and home domain name visible to the visited DA Home domain name required for PWC accounting SOLUTION: Use PWC privacy enhancement modules (PWC mixes)

10. 10 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC Mixes DA ‘P’ (Provider) DA ‘A’ (First mix) DA ‘B’ (Second mix) DA ‘C’ (Consumer) user_X@C “My PWC user ID is alias_X@A ” (Appends real ID and a mix chain, all encrypted using layered public-key encryptions) P, A, and B cannot know if the domain on the right is the real consuming domain or a mix A, B, and C cannot know if the domain on the left is the real providing (visited) domain or a mix Blue arrows represent token flow

11. 11 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr Open PWC Issues 1. Secure Distributed Accounting Maintains PWC accounting history Must be fault-tolerant, scalable, hack-proof 2. Tokens and Token Generation Cryptographically secure, unforgeable tokens Generated, perhaps, by a PWC internal distributed bank Distributed to new PWC entrants 3. Domain Heterogeneity Domains covering areas diverse in size and location Domains may have completely uneven populations of registered users Small domains may receive only very few requests (and thus tokens) 4. “Offline” Domains Domain Agent autonomy may mean a DA is unreachable/offline Who “pays” for a roaming user of that domain? The roaming user? Another domain?

12. 12 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr Deploying the PWC Domain Agent Administrative Interface Must hide PWC complexity from Domain Agent administrators DAs must require a minimum number of input parameters: 1. A list of registered users and their security credentials 2. The domain’s aggregate egress and ingress Internet bandwidth 3. A “map” of WLAN cells and local traffic bottlenecks 4. The average WLAN load (local registered users and visitors) 5. The average PWC usage by roaming users of the domain Some of these parameters will be administrator’s ‘best-guesses’ PWC Profit Opportunities 1. Vendors of PWC Domain Agents 2. Vendors of PWC support modules 3. PWC domain aggregators 4. “Pay-as-you-go” domains

13. 13 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr PWC Domain Agent Prototype We’ve built two prototype PWC Domain Agents Running on PCs with Red Hat Linux 9 (2.4.20 kernel) Developed using C, Java, and Python Each DA is also a WLAN router, connected to the Internet and to a Cisco Aironet 1200-series WLAN AP Modules completed: Authentication Using IEEE 802.1X Using a custom web-based login function (and the iptables firewall) Traffic Policing Using the libpcap library and the tc utility WLAN Using Linux IP masquerading (for NAT/NAPT) and standard Linux DHCP, DNS, and routing functionality Strategy (using a very simple P2P token-exchange rule) Still needed: Unforgeable tokens, secure DHT (for distributed accounting and name- service), more complex strategy algorithms, PWC mixes

14. 14 / 14 AUEB Mobile Multimedia Lab – http://mm.aueb.gr Concluding Remarks The PWC is a simple alternative to existing roaming schemes The PWC is designed around organic growth PWC strategic agents replace static roaming agreements Although, by design, the PWC cannot provide any strong guarantees, it could become a suitable vehicle for achieving ubiquitous, low-cost, Internet access PWC autonomy and privacy considerations could make it more socially acceptable Real-world regulations could, however, affect PWC growth More analysis and simulations are needed to assist in designing optimal PWC rules