Towards understanding, improving and securing BitTorrent protocol and system

Abstract

BitTorrent (BT) is one of the most common Peer-to-Peer (P2P) file sharing protocols. It was reported that the BT system has on average a day 20 millions users from 120 countries. The study of BT protocol and system has become an important aspect of P2P computing. Towards understanding, improving and securing BT protocol and system, we address several issues including modeling, algorithms and worms. 1) Modeling work provides insights into the BT system performance and the BT protocol improvement. Existing BT models are limited to a certain level peer level, cluster level, or swarm level. We build a bottom-up probabilistic model to analyze the BT system from the individual peer behavior to the overall system performance. 2) Piece-related algorithms play an inherently crucial role in the BT protocol, because the BT system relies upon peers to cooperatively trade their pieces with one another. Although existing algorithms were long believed to be good enough, our understanding of them is still far from complete. We propose a distributed credit method to prevent under-reporting, a utility-driven strategy to balance piece supply and demand, and a possible fix to address the conflict between piece selection and piece queuing. 3) P2P worms have become a serious threat to the Internet security, due to their fast and large-scale propagation in P2P systems. We present a novel Adaptive BitTorrent worm (A-BT worm) to discuss potential ways to design a powerful P2P worm. We then propose a hybrid model to estimate the worm damage, a statistical method to detect the worm behavior, and a safe strategy to slow down the worm propagation.