Traffic Logging in the Peer-to-Peer Wireless Network Confederation Pantelis Frangoudis, Elias C. Efstathiou, George C. Polyzos Department of Computer Science Athens University of Economics and Business Athens 104 34, Greece {pfrag, efstath, polyzos}@aueb.gr We have introduced the concept of a Peer-to-Peer Wireless Network Confederation (P2PWNC) in [1,2]. In this paper we present ongoing work towards implementing Traffic Logging in the context of a homogeneous P2PWNC that is WLAN-based. We first provide a brief overview of these concepts. The P2PWNC is a collection of wireless network domains. It is also a community in which users registered with any one of the participating domains can access services from any other participating domain within which they happen to roam. Assuming that the provided service is Internet access, then, the P2PWNC has the same goal as existing Wireless ISP roaming associations (see for example [3]). However, the P2PWNC is an exercise in Distributed Algorithmic Mechanism Design [4]. It belongs to a growing class of multi-agent systems (e.g. file-sharing networks) that reside at the intersection of Computer Science and Economics, in which both incentive compatibility and distributed computation issues arise. The following six characteristics make the P2PWNC distinct. (The economics of the P2PWNC are discussed in [5,6] while system implementation issues are raised in [7].) 1. Unlike the Hotspot Aggregation model (exemplified by [8,9]), in the P2PWNC there is no separate central authority with which all the mobile users need to register. 2. Unlike other roaming associations, there are no strict bilateral or multilateral peering agreements between domains, which enjoy significant autonomy concerning their sharing policies, i.e. the amount of resources they make available to visitors. 3. Simple rules that ensure fair participation and combat free-riding are used to guide a domain’s sharing policy. The rules specify the amount and type of resources that domains should make available to visitors. These rules are the subject of ongoing economic analysis [5]. 4. The P2PWNC domains generally operate in a reciprocal way: domains offer services to foreign visitors so that their own users may enjoy the same privilege when roaming. 5. The P2PWNC does not rely on any entities that are centralized or external to the P2PWNC for storing the community’s history: accounting information and domain rankings are stored in a decentralized and tamper-proof manner within the participating domains themselves. 6. The P2PWNC assumes no monetary exchanges and it is “free to use” (subject to community rules). The homogeneous, WLAN-based P2PWNC is the most basic type of P2PWNC: the participating domains are simple residential WLAN hotspots. Assume that broadband Internet subscribers install home WLAN routers that run P2PWNC software. With enough routers and because WLAN signals can (or be made to) extend outside, it is not difficult to imagine creating a blanket of coverage (at least in metropolitan areas) composed of thousands of cells operated by independent mini-providers. Although the same basic idea can be found in the business model of [10] none of the aforementioned six characteristics applies there. The P2PWNC WLAN routers, known as Domain Agents (or DAs – see Figure), allow other users that happen to be nearby to access the Internet, assuming these users can prove that they themselves are DA operators who follow the community’s sharing rules. Then, all the operators of such DAs could enjoy metropolitan wireless coverage at the cost of a single broadband ISP subscription. The main objective of this paper is to document the implementation issues that arise when attempting to enforce the P2PWNC participation rules using simple software that could run on a cheap, embedded device - the aforementioned WLAN routers that would double as DSL or cable modems. In this paper and for our prototype implementation, we concentrate on PC-based P2PWNC WLAN routers that run Linux. Our objective is to reuse as much as possible of the prototype’s code in the embedded version. The rule we are currently trying to enforce is a volume-based one: the more visitor traffic a DA routes, the more participation “points” it receives. Visitors “pay” using some of their home domain’s points. Traffic volume is measured both at the IP and at the application layer, meaning that data units such as e-mails and web pages, or sessions like SSH, Telnet, and FTP are also taken into account. If the mobile visitor makes use of IPSEC or other end-to-end encryptions, DAs fall back to measuring bulk IP volume only.