Обсуждение: [HACKERS] Authentification method on client side checking

Hello,

Despite the addition of SCRAM authentification to PostgreSQL 10, MITM 
attack can be performed by saying that the server supports, for example, 
only md5 authentication. The possible solution for it is checking 
authentification method on a client side and reject connections that 
could be unsafe.

Postgresql server can require unencrypted password passing, md5, scram, 
gss or sspi authentification.

In the attached patch you can find the solution for it. The new provided 
features are the following:
The parameter with acceptable authentification methods can be passed 
into connection methods of libpq library.
Also, this parameter can be specified to psql as a command line 
argument.
The documentation for command line arguments of psql and arguments of 
libpq methods are also presented.

Thank you for attention!

Best,
-- 
------
Victor Drobny
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company
-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

On 09/07/17 18:47, Victor Drobny wrote:
> Hello,
>
> Despite the addition of SCRAM authentification to PostgreSQL 10, MITM 
> attack can be performed by saying that the server supports, for 
> example, only md5 authentication. The possible solution for it is 
> checking authentification method on a client side and reject 
> connections that could be unsafe.
>
> Postgresql server can require unencrypted password passing, md5, 
> scram, gss or sspi authentification.
    Hi Victor.
    Precisely yesterday I initiated a similar thread: 
https://www.postgresql.org/message-id/d4098ef4-2910-c8bf-f1e3-f178ba77c381%408kdata.com
    I think that a) the mere auth mechanism is not enough (channel 
binding or not, ssl or not, change a lot the effective security 
obtained) and b) maybe a categorization is a better way of specifying a 
connection security requirements.
    What's your opinion on this? Any answer should also be coordinated 
among the drivers.

    Álvaro


-- 

Álvaro Hernández Tortosa


-----------
<8K>data

On Mon, Jul 10, 2017 at 9:29 AM, Álvaro Hernández Tortosa
<aht@8kdata.com> wrote:
>     Precisely yesterday I initiated a similar thread:
> https://www.postgresql.org/message-id/d4098ef4-2910-c8bf-f1e3-f178ba77c381%408kdata.com
>
>     I think that a) the mere auth mechanism is not enough (channel binding
> or not, ssl or not, change a lot the effective security obtained) and b)
> maybe a categorization is a better way of specifying a connection security
> requirements.
>
>     What's your opinion on this? Any answer should also be coordinated among
> the drivers.

Before rushing into implementing something that we may not want, let's
discuss the matter on the thread spawned by Álvaro and find an
agreement and a direction of implementation. I was planning to answer
your message with my own thoughts on the matter. Having more control
in libpq is definitely something that we should have.
--
Michael