readme.ssl

来自「PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统」· SSL 代码 · 共 484 行 · 第 1/2 页

to clients (via certificates), but clients can remain anonymous.
This protects clients from "man-in-the-middle" attacks (where a
bogus server either captures all data or provides bogus data),
but does not protect the server from bad data injected by false
clients.

The community has established a set of criteria for secure
communications:

*) the server must provide a certificate identifying itself
   via its own fully qualified domain name (FDQN) in the
   certificate's Common Name (CN) field.

*) the FQDN in the server certificate must resolve to the
   IP address used in the connection.

*) the certificate must be valid.  (The current date must be
   no earlier than the 'notBefore' date, and no later than the
   'notAfter' date.)

*) the server certificate must be signed by an issuer certificate
   known to the clients.

   This issuer can be a known public CA (e.g., Verisign), a locally
   generated root cert installed with the database client, or the 
   self-signed server cert installed with the database client.

   Another approach (used by SSH and most web clients) is for the
   client to prompt the user whether to accept a new root cert when
   it is encountered for the first time.  psql(1) does not currently
   support this mechanism.

*) the client *should* check the issuer's Certificate Revocation
   List (CRL) to verify that the server's certificate hasn't been
   revoked for some reason, but in practice this step is often
   skipped.

*) the server private key must be owned by the database process
   and not world-accessible.  It is recommended that the server
   key be encrypted, but it is not required if necessary for the
   operation of the system.  (Primarily to allow automatic restarts
   after the system is rebooted.)
  
The 'mkcert.sh' script can be used to generate and install 
suitable certificates

Finally, the client library can have one or more trusted root
certificates compiled into it.  This allows clients to verify
certificates without the need for local copies.  To do this,
the source file src/interfaces/libpq/fe-ssl.c must be edited
and the database recompiled.

Secure Channel: Mutual Authentication
-------------------------------------

Mutual authentication requires that servers and clients each
authenticate to the other.  This protects the server from
false clients in addition to protecting the clients from false
servers.

The community has established a set of criteria for client
authentication similar to the list above.

*) the client must provide a certificate identifying itself.
   The certificate's Common Name (CN) field should contain the
   client's usual name.

*) the client certificate must be signed by a certificate known
   to the server.

   If a local root cert was used to sign the server's cert, the
   client certs can be signed by the issuer.

*) the certificate must be valid.  (The current date must be
   no earlier than the 'notBefore' date, and no later than the
   'notAfter' date.)

*) the server *should* check the issuer's Certificate Revocation
   List (CRL) to verify that the clients's certificate hasn't been
   revoked for some reason, but in practice this step is often
   skipped.

*) the client's private key must be owned by the client process
   and not world-accessible.  It is recommended that the client
   key be encrypted, but because of technical reasons in the
   architecture of the client library this is not yet supported.

PostgreSQL can generate client certificates via a four-step process.

1. The "client.conf" file must be copied from the server.  Certificates
   can be highly localizable, and this file contains information that
   will be needed later.

   The client.conf file is normally installed in /etc/postgresql/root.crt.
   The client should also copy the server's root.crt file to
   ~/.postgresql/root.crt.

2. If the user has the OpenSSL applications installed, they can
   run pgkeygen.sh.  (An equivalent compiled program will be available
   in the future.)  They should provide a copy of the
   ~/.postgresql/postgresql.pem file to their DBA.

3. The DBA should sign this file the OpenSSL applications:

     $ openssl ca -config root.conf -ss_cert ....

   and return the signed cert (postgresql.crt) to the user.

4. The user should install this file in ~/.postgresql/postgresql.crt.

The server will log every time a client certificate has been
used, but there is not yet a mechanism provided for using client
certificates as PostgreSQL authentication at the application level.


---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/users-lounge/docs/faq.html

------------------------------------------------------------------------------

Date: Tue, 21 May 2002 19:50:38 -0400
From: Neil Conway <[e-mail]>
To: "Bear Giles" <[e-mail]>
cc: [e-mail]
Subject: Re: [HACKERS] 2nd cut at SSL documentation

On Tue, 21 May 2002 14:27:00 -0600 (MDT)
"Bear Giles" <[e-mail]> wrote:
> A second cut at SSL documentation....

I've pointed out some minor things I noticed while reading through.
Yeah, I was bored :-)

> The sites that require SSL fall into one (or more) of several broad
> categories.
> 
> *) They have insecure networks. 
> 
>    Examples of insecure networks are anyone in a "corporate hotel,"

What's a corporate hotel?

> *) They have 'road warriors.'

This section title sounds confusingly similar to the 1st item.
Perhaps "They need to authentication clients securely" or something
similar? The need to use client certificates does not apply only to
"road warriors" -- I've seen situations where client-certs are used for
clients to connecting to a server over a LAN.

> *) Access is limited to the Unix socket.

"the" sounds wrong, there's more than just 1 :-)

> *) Access is limited to a physically secure network.
> 
>    "Physically secure" networks are common in the clusters and
>    colocation sites - all database traffic is restricted to dedicated
>    NIC cards and hubs, and all servers and cabling are maintained in
>    locked cabinets.

Perhaps add a note on the performance hit here?

> Using SSH/OpenSSH as a Virtual Private Network (VPN)

I'm unsure why you're bothering to differentiate between SSH
and OpenSSH.

> SSH and OpenSSH can be used to construct a Virtual Private Network
> (VPN)

No need to include the abbreviation for VPN here, you've explained
the term before.

> to provide confidentiality of PostgreSQL communications.  
> These tunnels are widely available and fairly well understood, but 
> do not provide any application-level authentication information.

You might want to clarify what "application-level authentication
information" means, or else leave out all discussion of drawbacks
until later.

> To set up a SSH/OpenSSH tunnel, a shell account for each
> user should be set up on the database server.  It is acceptable
> for the shell program to be bogus (e.g., /bin/false), if the
> tunnel is set up in to avoid launching a remote shell.
> 
> On each client system the ~/.ssh/config file should contain
> an additional line similiar to
> 
>  LocalForward 5555 psql.example.com:5432

"pgsql.example.com" strikes me as a better example hostname (I always
think that psql == DB client, postgres/postmaster/pgsql == DB server).

> Unfortunately, there are many potential drawbacks to SSL tunnels:

I think you mean SSH tunnels.

> *) the SSH implementation or protocol may be flawed.  Serious problems
>    are discovered about once every 18- to 24- months.

I'd be skeptical whether this weakness if specific to SSH -- there
can be security holes in OpenSSL, the SSL protocol, the SSL
implementation in PostgreSQL, etc.

> *) the database server must provide shell accounts for all users
>    needing access.  This can be a chore to maintain, esp. in if

Remove the "in".

> *) neither the front- or back-end can determine the level of
>    encryption provided by the SSH tunnel - or even whether an
>    SSH tunnel is in use.  This prevents security-aware clients
>    from refusing any connection with unacceptly weak encryption.

Spelling error.

> Finally, the client library can have one or more trusted root
> certificates compiled into it.  This allows clients to verify
> certificates without the need for local copies.  To do this,
> the source file src/interfaces/libpq/fe-ssl.c must be edited
> and the database recompiled.

"PostgreSQL" recompiled -- database versus RDBMS can be ambiguous.

> Mutual authentication requires that servers and clients each
> authenticate to the other.  This protects the server from
> false clients in addition to protecting the clients from false
> servers.

"false" in this context?

Cheers,

Neil

-- 
Neil Conway <[e-mail]>