[SR-Users] Obscuring SIP traffic and using with NoSIP

[Daniel-Constantin Mierla]

On 30/07/14 06:37, Muhammad Shahzad wrote:
> Humm, no reply so far, may be because my email was very long and no 
> body bothered to read it all. Anyways, here is the shorter more direct 
> version of it. (including kamailio dev list, since question is rather 
> technical).
>
> Is it possible to implement a custom SIP transport in Kamailio script 
> file i.e. kamailio.cfg. The purpose is to allow experimentation with 
> custom encryption algorithms such as this,
>
> https://github.com/mshary/itv
>
> What we need is a couple of functions, one to receive incoming raw / 
> encrypted data received on SIP socket, which then can be parsed / 
> decrypted in kamailio.cfg (using e.g. LUA or PERL language modules 
> etc.) and afterwords feed to kamailio for usual processing (as if it 
> was normal / plain-text sip data received on sip socket). The second 
> function to do the opposite, it receives the normal / plain-text sip 
> data that is ready to be sent out from kamailio's core, encrypts it 
> and then send it out to actual destination.
>
> In case above is not possible. Can i do it in kamailio's native code? 
> Any hooks / example code for reference?
If you look at encrypting sip messages, look at topoh module. You can 
write a replacement for its hooks. Topoh is practically decoding the 
headers and then lets the pure SIP message go through config file 
execution. Before sending, it encodes the headers and then let it go to 
the network.

This is something that should be rather straightforward to do if you are 
familiar with C code.

You mentioned that using TLS can still reveal patters of being sip. You 
have to think here of ways to obfuscate even in your case of a new 
encryption method. What can be matched here:
- periodical registrations - you can have the client (or even the 
server) to use different expires times for each registration
- size of packages, specially if user IDs are the same or similar length 
(e.g., say everyone uses a 10 digit id), practically no matter who is 
calling who, the size will be pretty much the same because most of the 
phones I have seen so far use same set of headers. Here you can add 
random custom headers for each packet. I haven't checked the proposed 
encryption algorithm (some use random blocks implicitly to pad the 
data), but eventually you can add random data before and after the 
packet that you strip (and re-add) in topoh-replacement module

The other option of having a totally different protocol than SIP should 
be possible as well. But you need to re-implement a lot (like location, 
authentication, ...). Look at msrp module for an example. This may need 
to touch core code a bit.

Of course, in both cases, the client application has to be developed as 
well. Perhaps still easier if going for first option, by reusing some 
open source sip client and adding the encapsulation/decapsulation layer 
when receiving/sending to network.

Cheers,
Daniel

>
> Many thanks and kind regards for your help.
>
>
> On Mon, Jul 28, 2014 at 2:38 AM, Muhammad Shahzad 
> <shaheryarkh at gmail.com <mailto:shaheryarkh at gmail.com>> wrote:
>
>     Hi,
>
>     As the mobile voip is getting more and more popular these days,
>     there has been a strong opposition from GSM operators against
>     mobile voip apps. They often use tactics like blocking voip ports,
>     or detect and block voip traffic and in some cases restricting udp
>     traffic altogether to very low upload and download speeds. See
>     below link for some details,
>
>     http://www.linphone.org/eng/blog/linphone-over-3g.html
>
>     While not all the problems can be solved right now (especially the
>     limiting udp traffic, since RTP always uses udp transport) I was
>     wondering if we can at least handle the sip related problems. The
>     most important of them is SIP traffic detection. While some forks
>     would suggest using TCP/TLS to encrypt SIP traffic, it has a few
>     problems, e.g.
>
>     1. It requires somewhat high resources on mobile devices, so many
>     low-end android phones simply can't use it.
>
>     2. There is possibility that encryption signature may identify it
>     as SIP traffic. There exists firewalls (often deployed in middle
>     eastern countries) which have huge database of encryption
>     signatures and patterns which although may not decrypt the sip
>     packet but at least identify it as sip packet and block it.
>
>     Also with rough agencies of evil empires spying over millions of
>     users worldwide makes the current encryption standards pretty much
>     pointless, at least in terms of user privacy and network security.
>     So there is a strong need to experiment with new ideas and
>     concepts to regain internet freedom. Some of such ideas are,
>
>     1. Convert sip traffic which is plain text to binary format just
>     before transmitting it and revert it to plain text upon reception.
>
>     2. XOR the sip traffic (pretty much same as binary sip).
>
>     3. Use some very lightweight but effective / non-standard
>     encryption algorithm, e.g.
>
>     https://github.com/mshary/itv
>
>     All these ideas require that SIP server such as Kamailio is able
>     to adopt to these, preferably with minimal or no change in native
>     code. The NoSIP module seems an interesting module in this regard.
>     It provides all traffic it thinks is not the SIP traffic to
>     configuration script, where we can do our own parsing and do
>     whatever we want with it. I have two questions about this,
>
>     1. If parsed message is SIP, we can we send it back to kamailio
>     core to get it processed as if it is a normal SIP message received
>     by kamailio?
>
>     2. Can this module or any other module available in kamailio, that
>     can provide us full sip packet that is about to be transmitted
>     over sip socket, so we can "encode" it just before it is sent to
>     next hop?
>
>     I know this would be like writing a SIP transport in kamailio
>     script which would be very tough if not impossible to implement in
>     native core. But it will really help in winning the modern mobile
>     voip challenges.
>
>     Thank you.
>
>
>

-- 
Daniel-Constantin Mierla - http://www.asipto.com
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-users/attachments/20140730/257a79c9/attachment.html>