Skip to content
Snippets Groups Projects
Select Git revision
  • experimental
  • main default protected
2 results
Compare
user avatar
Change how the length obfsucation mask is derived.
Yawning Angel authored 
Instead of using the nonce for the secret box, just use SipHash-2-4 in
OFB mode instead.  The IV is generated as part of the KDF.  This
simplifies the code a decent amount and also is better on the off
chance that SipHash-2-4 does not avalanche as well as it is currently
assumed.

While here, also decouple the fact that *this implementation* of obfs4
uses a PRNG with 24 bytes of internal state for protocol polymorphism
instead of 32 bytes (that the spec requires).

THIS CHANGE BREAKS WIRE PROTCOL COMPATIBILITY.
5bdc376e
History
user avatar 5bdc376e
History
Name Last commit Last update
csrand
doc
drbg
framing
ntor
obfs4proxy
.gitignore
README.md
handshake_ntor.go
handshake_ntor_test.go
obfs4.go
packet.go
replay_filter.go
replay_filter_test.go
weighted_dist.go
README.md

obfs4 - The obfourscator

Yawning Angel (yawning at torproject dot org)

WARNING

This is pre-alpha. Don't expect any security or wire protocol stability yet. If you want to use something like this, you should currently probably be looking at ScrambleSuit.

What?

This is a look-like nothing obfuscation protocol that incorporates ideas and concepts from Philipp Winter's ScrambleSuit protocol. The obfs naming was chosen primarily because it was shorter, in terms of protocol ancestery obfs4 is much closer to ScrambleSuit than obfs2/obfs3.

The notable differences between ScrambleSuit and obfs4:

  • The handshake always does a full key exchange (no such thing as a Session Ticket Handshake).
  • The handshake uses the Tor Project's ntor handshake with public keys obfuscated via the Elligator 2 mapping.
  • The link layer encryption uses NaCl secret boxes (Poly1305/XSalsa20).

Why not extend ScrambleSuit?

It's my protocol and I'll obfuscate if I want to.

Since a lot of the changes are to the handshaking process, it didn't make sense to extend ScrambleSuit as writing a server implementation that supported both handshake variants without being obscenely slow is non-trivial.

TODO

  • Code cleanups.
  • Write more unit tests.
  • Optimize further.

WON'T DO

  • I do not care that much about standalone mode. Patches MAY be accepted, especially if they are clean and are useful to Tor users.
  • Yes, I use a bunch of code from the borg^w^wGoogle. If that bothers you feel free to write your own implementation.
  • I do not care about older versions of the go runtime.

Thanks

  • David Fifield for goptlib.
  • Adam Langley for his Elligator implementation.
  • Philipp Winter for the ScrambleSuit protocol which provided much of the design.