The security of the Internet's interdomain routing system hinges on whether autonomous systems (ASes) can trust the information they receive from each other via the Border Gateway Protocol (BGP). Frequently, this trust has been misguided, resulting in wide-spread outages and significant concerns about future attacks. Despite the seriousness of these problems, proposals for a more secure version of BGP have been stymied by serious impediments to practical deployment. Instead, we argue that the existing trust relationships between network operators (and the institutions they represent) are a powerful force for improving the security of BGP, without changing the underlying routing protocol. Our approach leverages ideas from online reputation systems to allow ASes to form a peer-to-peer overlay that integrates results from local network-management tools for detecting attacks and configuration errors. The proposed architecture is incrementally deployable, protects against shilling attacks, and deters malicious operator behavior.