We present a flexible framework for throttling attackers in peer-to-peer media streaming systems. In such systems, selfish nodes (e.g., free riders) and malicious nodes (e.g., DoS attackers) can overwhelm the system by issuing too many requests in a short interval of time. Since peer-to-peer systems are decentralized, it is difficult for individual peers to limit the aggregate download bandwidth consumed by other remote peers. This could potentially allow selfish and malicious peers to exhaust the system's available upload bandwidth. In this paper, we propose a framework to provide a solution to this problem by utilizing a subset of trusted peers (called kantoku nodes) that collectively monitor the bandwidth usage of untrusted peers in the system and throttle attackers. This framework has been evaluated through simulation thus far. Experiments with a full implementation on a network testbed are part of our future work.