This thesis focuses on the problem of tracking illegal P2P content distributors. By viewing the collection of files of a peer as a relatively precise reflection of its owner, we use the Ising model which originates from statistical physics to mathematically model the behavior of P2P networks and identify the relationships of peers. Based on it, we develop an effective approach to track the behavioral-based structures of P2P networks and use it as a guidance to narrow down the search scope for illegal P2P content distributors. The sum-product algorithm and mean field algorithm which are based on the Ising model are then used to efficiently compute the marginal distribution of peers that are holding or held a particular file of known contraband. Experimental results have shown that this behavioral-based approach significantly outperforms several tracking algorithms that ignore the relationships of peers in P2P networks.