More contest from the tor-dev announcement. Not sure why they don't have the paper in the repo:
The latest version of our paper is available on the Cryptology ePrint
Archive . In the paper we describe our framework and provide a
comprehensive security evaluation of FTE's success in evading six DPI
systems --- including using regular expressions from open-source DPI
systems to evade detection by a closed-source black-box commercial DPI
system. By "evade" we mean that it's easy for FTE to tunnel arbitrary
TCP streams (e.g., Tor) such that they are (mis)classified by DPI
systems as a configurable target protocol (e.g., HTTP, SMB, RTSP,
etc.) of one's choosing.
We release FTE in its alpha stage because we believe it is well
positioned to evade the suspected protocol white-listing  recently
reported in Iran. More generally, we're optimistic FTE has long-term
potential as a tool to enable users to control how their traffic is
classified by passive DPI systems. As one example, over the last
month, we've successfully tunneled Tor through the Great Firewall of
China, using FTE to make our traffic "look like" HTTP.
We're eager for feedback on this alpha release, so please do not
hesitate to contact us with questions.