Boycotting and Extorting Nodes in an Internetwork

A boycott is a protest or a demand for change based on consumers following a simple rule: do not purchase service or goods from specific producers. In any network, and especially in the Internet, such grass-roots protest is close to impossible. The big networks do not connect directly to consumers and routing is based on locally choosing the shortest, cheapest, or most profitable paths. In this environment, individual consumers have no influence on route selection and enterprise customers very little. Yet, protest-worthy actions by network providers are not infrequent. AOL compromised the privacy of its users by publishing search queries (August 2006); Google complied with China’s censorship laws, to the disapproval of many (February 2006); Verisign redirected mistyped DNS names to their own advertising (September 2003); and the Electronic Frontier Foundation has sued AT&T to stop NSA surveillance (January 2006; legal action continues). We envision an Internet in which users have the ability to make, or at least influence, routing decisions. Users will have this ability so that they may take advantage of increasingly rich functionality in networks; users will need to be able to choose networks that provide a service or avoid those that filter or censor. Among the many policies users might choose in routing packets, users may select nodes to boycott. To study the economics of boycotting a provider in an abstract network, we adapt the Vickrey-Clarke-Groves (VCG) routing mechanism to support boycotting. All link costs are globally advertised. Sources in this network typically choose the shortest cost path to a destination, and micropayments reward each entity along the path. Each of these assumptions is distinct from the realities of BGP; we describe the implications of this mismatch late in this paper. We begin the investigation into the following questions: (1) Does VCG continue to encourage nodes to report costs truthfully in the presence of boycotting? (2) Do nodes gain from learning users’ boycott lists? Do users gain from divulging whom they are boycotting? and (3) How do the results from VCG apply to ad hoc networks and the Internet? To answer these questions, we modify VCG (§2) and show that only when massively boycotted do nodes have incentive to lie about link costs (§3). We simulate our modified VCG (§4) to show the price in connectivity, and discuss deployment requirements and implications (§5). We review related work (§6) and conclude (§7). 2. A MECHANISM FOR BOYCOTTING 2.1 Model, Assumptions, and Problem