There is contextual data that can help point toward “possible” (cough, “probable”) infringing content distribution. http://ha.ckers.org/blog/20091008/css-history-hack-used-to-ban-torrent-users/ is an example. But I doubt if it is the “tip of the iceberg”. When non-infringing content distribution deals turn toward torrents to save on network distribution costs, they’ll need a way to tell the difference between them and the rest.
Legal processes might favor solutions like creating “distribution registries” with a goal to disambiguate non-infringing content distribution. It would have to work with other assisting technologies.