My post is called "The great Kamailio security debate and some misconceptions debunked" and can be read here:
The ToC:
- Introduction
- A bit of background before diving in
- Claim: this issue does not affect many organisations
- Claim: custom headers are only known to internal users
- Claim: if it’s an 18 year old bug, it can’t have been high risk
- Claim: this should have been found if people were doing proper testing
- Claim: infrequent advisories = project is not serious about security
- Claim: limited number of advisories = project is more secure
- Claim: if you’re serious about security, monitor the mailing lists
- Claim: security experts should decide what is a security vulnerability
- Discussion: when should the project publish an advisory?
- Discussion: educational security role
- Moving forward
Hope that it is at least interesting and perhaps even constructive!
Best wishes,
--
Sandro Gauci, CEO at Enable Security GmbH
Register of Companies: AG Charlottenburg HRB 173016 B
Company HQ: Pappelallee 78/79, 10437 Berlin, Germany
On Thu, 3 Sep 2020, at 10:34 AM, Olle E. Johansson wrote:
Well, you have defined one definitive line between being stupid and following some current practise :-)
I like to think we as a project have an educational role as well. In this case explaining the bug we had and what it can cause.
We should definitely add a warning along the lines you write too - relying on headers alone is bad and not best current practise.
/O
After 20 years in voip, my 2 cents on this, if you succeed in creating a voip system where the security of the whole relies on the ability to remove (or only keep specific) custom sip headers, you will wake up one morning realizing a bunch of people in Palestine made a gazillion calls over your system to expensive destinations, bringing you to or over the edge of bankruptcy.
Security should be multilayered, one header sneaking through should not give any big problems.
From a security point of view, this could be called a 'normal' security risk, I think. It's a bit more than low as you can do more than just get some info, but it's not high, as you would need to have many other factors going wrong to get to a successful exploit.
Op do 3 sep. 2020 om 09:18 schreef Olle E. Johansson <
oej@edvina.net>:
One thought - we may have to separate security vulnerability reporting from security advisory documents. I think in some cases, where a common use of a product can lead to issues (but it is not clearly a bug that cause crashes in our code) we may have to send out an advisory and publish it in the same way. The problem with that is where the border is between just doing stupid things like taking SQL statements from SIP headers and issues like this that are harder to catch.
We had a long and hard discussion about this in the Asterisk project many years ago - a very common dialplan construct (that was documented in many places) was indeed very dangerous. It wasn’t any code in asterisk that caused the issue, just a common dialplan construct that existed in many, many production systems. In the end, if I remember correctly, the project issued an advisory and added a README about it.
Maybe that’s a way forward.
/O
Hello Maxim,
have a look to the first sentence:
“A security vulnerability is (for example) when a user of Kamailio can cause Kamailio to crash or lock up by sending messages to the server process.”
So there is some limitation regarding vulnerability criticality defined in there. But of course (as I already mentioned), it might be improved to e.g. use CVSS scoring instead.
Cheers,
Henning
On Wed, Sep 2, 2020 at 11:30 AM Henning Westerholt <
hw@skalatan.de> wrote:
Hello Maxim,
thank you for the clarification, appreciated.
No worries, hope to have a civilized discussion.
Just one clarification, my comment regarding the advisory from 2018 was not meant as advertisement etc..
Point taken, I dramatized of course to underline my point.
One suggestion to objectify the whole discussion, there exists a well-known and accepted metric for vulnerabilities: CVSS [1]
If I calculate the CVSS score for this issue, it results in a medium level with score 5.8. But this is of course again (at least somewhat) influenced from my point of view to this bug.
Some projects have a policy to only do a security announcement for vulnerabilities with score high and critical. For Kamailio this is not yet defined in a detailed way, due to the size of the project and other factors.
So, If people in this discussion (or other people on the list) are interested in improving the project security processes – this wiki page with the current process might be a good starting point:https://www.kamailio.org/wiki/security/policy
Please suggest your improvements to the existing process (preferable in a new discussion thread) on the sr-dev list. If you want to do it in private, feel free contact the management list.
Well, first suggestion after having read it: to start actually following what's documented before any improvements are made. ;-) The policy says plain and simple (quote):
Publishing security vulnerabilities
Kamailio will publish security vulnerabilities, including an CVE ID, on the kamailio-business mailing list, sr-dev, sr-users as well as related lists. The advisories will also be published on the kamailio.org web site. CVE entries should be created for vulnerabilities in the core and major modules, for rarely used modules this is not necessary. If there are several security issues together in one release, they should be announced together.
I might be missing something obvious, but there is no "if" or "maybe" or "it depends". Any module that has been 18 years with the project qualifies to be a "major module" to me...
_______________________________________________
Kamailio (SER) - Users Mailing List
_______________________________________________
Kamailio (SER) - Users Mailing List
_______________________________________________
Kamailio (SER) - Users Mailing List
_______________________________________________Kamailio (SER) - Users Mailing Listsr-users@lists.kamailio.orghttps://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users