Re: [mod-security-users] New SecRemoteAddrDefine (httpd-dev CCed)
Brought to you by:
victorhora,
zimmerletw
Menu
▾
▴
Re: [mod-security-users] New SecRemoteAddrDefine (httpd-dev CCed)
|
From: Breno S. <bre...@gm...> - 2013-05-06 14:00:05
|
I think SecDefineRemoteAddr should be present. People just need to understand that it will only change the IP address for mod_security and not for all apache modules. If people are using Apache 2.4 + mod_remoteip or Apache 2.2 + rpaf, just do not set SecDefineRemoteAddr. If no modules like this are present in Apache / Nginx / IIS people will have at least one option builtin in mod_security to get the client ip address On Mon, May 6, 2013 at 10:49 AM, Reindl Harald <h.r...@th...>wrote: > Am 06.05.2013 15:20, schrieb Breno Silva: > > * in this case they also would not configure SecRemoteAddrDefine > > Why not ? If the proxy/load balancer is setting the X-Forwarded-For we > could extract the data using > > SecDefineRemoteAddr right ? > > but you do you expect people not care about the implications of a > reverse-proxy/load-balancer start to use SecRemoteAddrDefine correctly > > if they would do things right they would use rpaf or whatever > module/solution > with the same functionality and not need SecRemoteAddrDefine > > even if they do - for what benefit? > in the overall context mod_security is only a small piece > _____________________________ > > there are three options: > > * care about how to do it right and find out that mod_security > is a showstopper to the the right behavior in any layer of > your infrastructure as i found out with < 2.7.4 > > * care not at all, in this case SecRemoteAddrDefine would not get > configured > > * care partly without understand what you are doing, in this case > you start using SecRemoteAddrDefine because you found it somewhere > and the worst case is that people think they are done after that > while the app-developer happily continues to code > if(substr($_SERVER['REMOTE_ADDR'], '10.0.0.') !== false) not realizing > that all connections coming from 10.0.0.x > _____________________________ > > if you think it is a good idea, go ahead but please be careful > not to introduce regressions somewehere in mod_security because > this is a really sensitive place and mistakes can do a lot of harm > > > On Mon, May 6, 2013 at 10:17 AM, Reindl Harald <h.r...@th...<mailto: > h.r...@th...>> wrote: > > > > > > Am 06.05.2013 15:11, schrieb Breno Silva: > > > Yes.. but we cannot assume all users is doing it right :) > > > And to be honest i think many are not doing it. > > > > in this case they also would not configure SecRemoteAddrDefine > > > > > Do you have a box without rpaf or you can disable it to test > SecDefineRemoteAddr ? > > > > no, all the development stuff is on Apache 2.4 since 2013/01 and > > with mod_security 2.7.4 production servers will also be upgraded > > > > > Also i will need this feature for nginx/iis module. > > > > in my opinion this is clearly up to the webserver and not > > a random extension, it maybe does more harm as it helps > > because the overall behavior get's unpredictable > > > > however, please do not forget revisit "Fixed mod_security displaying > wrong ip > > address in error.log using apache 2.4" from modsec 2.7.2! > > > > > On Mon, May 6, 2013 at 10:08 AM, Reindl Harald < > h.r...@th... <mailto:h.r...@th...> > > <mailto:h.r...@th... <mailto:h.r...@th...>>> > wrote: > > > > > > i do not think so > > > > > > anybody which is running his webserver behind a load-balancer > > > without a solution like rpaf is obviously a fool because any > > > apache error/access-log is useless, any Allow/Deny does not > > > work as expected and last but not least REMOTE_ADDR in CGI > > > and PHP scripts is the address from the proxy and god beware > > > you have restrictions based on the client IP in your scripts > > > and not aware that you always have the same client-IP > > > > > > Am 06.05.2013 15:04, schrieb Breno Silva: > > > > Good. But is think we still need SecDefineRemoteAddr for > Apache 2.2 without rpaf right ? > > > > > > > > On Mon, May 6, 2013 at 10:02 AM, Reindl Harald < > h.r...@th... <mailto:h.r...@th...> > > <mailto:h.r...@th... <mailto:h.r...@th...>> > > > <mailto:h.r...@th... <mailto:h.r...@th...> > <mailto:h.r...@th... > > <mailto:h.r...@th...>>>> wrote: > > > > > > > > thank you, this works exactly as expected with Apache > 2.4 and > > > > mod_remoteip / mod_security, how i tested is expplained > at bottom > > > > > > > > PLEASE revisit the mod_security 2.7.2 change > > > > * Fixed mod_security displaying wrong ip address in > error.log using apache 2.4 and mod_remoteip > > > > > > > > this was obviously a wrong intented hack only affacted > the logging > > > > and without look at the source i guess you are logging > "X-Forwarded-For" > > > > in case it exists without respect mod_remoteip > > > > > _________________________________________________________________ > > > > > > > > SecRemoteAddrDefine is not needed beause with Apache 2.2 > and mod_rpaf > > > > this works all the time caused by a different handling > inside Apache > > > > and you need "mod_rpaf" there at least for the correct > ip-address > > > > in the accesslog to have useable webalizer-stats > > > > > > > > even with Apache 2.2 mod_security should not re-invent > the wheel > > > > it's up to Apache to handle this correctly and it does if > > > > correctly configured > > > > > _________________________________________________________________ > > > > > > > > and to answer Ryan Barnett's "What is preventing an > attacker from forging > > > > fake x-forwarded-for headers while still sending the > requests through > > > > a "trusted" proxy?" here the a snippet form the > httpd-docs > > > > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#processing > > > > > When multiple, comma delimited useragent IP addresses > are listed in > > > > > the header value, they are processed in Right-to-Left > order. Processing > > > > > halts when a given useragent IP address is not trusted > to present the > > > > > preceding IP address. The header field is updated to > this remaining > > > > > list of unconfirmed IP addresses, or if all IP > addresses were trusted, > > > > > this header is removed from the request altogether > > > > > _________________________________________________________________ > > > > > > > > testing to prove the correct working of the attached > source from you > > > > > > > > * UserAgent Switcher and my Firefox claims to be "Nessus" > > > > * SecRule REMOTE_ADDR "^10\.0\.0\.99" > "id:'102',phase:1,pass,nolog,ctl:ruleRemoveById=990002" > > > > * 10.0.0.103 is the real IP of the Apache-Trafficserver > > > > * with the configuration below i get no access denied as > expected > > > > * if i replace the "10.0.0.103" with "10.0.0.104" and > restart httpd > > > > as expected i get access denied > > > > * that's the state which i expected and demanded since > january > > > > > > > > <IfVersion >= 2.4> > > > > LoadModule remoteip_module > "modules/mod_remoteip.so" > > > > RemoteIPHeader X-Forwarded-For > > > > RemoteIPInternalProxy 127.0.0.1 10.0.0.103 > > > > RemoteIPProxiesHeader X-Forwarded-For > > > > </IfVersion> > > > > > _________________________________________________________________ > > > > > > > > Am 06.05.2013 14:30, schrieb Breno Silva: > > > > > Let's try this patch. Should work for Apache 2.4 + > mod_remoteip and Apache2.2 with the > > > SecDefineRemoteAddr > > > > > > > > > > On Mon, May 6, 2013 at 9:19 AM, Reindl Harald < > h.r...@th... > > <mailto:h.r...@th...> <mailto:h.r...@th...<mailto: > h.r...@th...>> > > > <mailto:h.r...@th... <mailto:h.r...@th...> > <mailto:h.r...@th... > > <mailto:h.r...@th...>>> > > > > <mailto:h.r...@th... <mailto: > h.r...@th...> <mailto:h.r...@th... > > <mailto:h.r...@th...>> <mailto:h.r...@th...<mailto: > h.r...@th...> > > > <mailto:h.r...@th... <mailto:h.r...@th...>>>>> > wrote: > > > > > > > > > > why do you refuse to understand that we do not > need a new feature or at > > > > > it is NOT up to you to re-invent the wheel > > > > > > > > > > Apache 2.2 works with proxy and mod_rpaf does not > need any change > > > > > in mod_security, mod_rpaf does the same as > mod_remoteip in a different > > > > > way which is no longer the way to go with Apache > >= 2.4 > > > > > > > > > > if i use %a in log-configuration i have ALWAYS the > desired behavior > > > > > with Apache 2.2 with and without proxy AND Apache > 2.4 with and without > > > > > mod_remoteip as i am currently run Apache 2.2 with > mod_rpaf which > > > > > everybody in context of load-balancers does and so > you do not need to > > > > > care about this in mod_security because in Apache > 2.2 %h contains the > > > > > correct address and in Apache 2.4 %a does the same > > > > > > > > > > until a few weeks ago nobody cared about this at > all > > > > > > > > > > now if it is brought up by me that Apache 2.4 has > BUILTIN support for > > > > > proxy-handling of the remote-addr simply respect > this in case we are > > > > > running under Apache 2.4 and understand that it is > the completly wrong > > > > > way you are going all the time starting with > hacking the modsec-logging > > > > > > > > > > > http://httpd.apache.org/docs/current/mod/mod_log_config.html > > > > > %a Client IP address and port of the request. > > > > > %{c}a Underlying peer IP address and port of the > connection (see the mod_remoteip module) > > > > > > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#remoteipheader > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#remoteipinternalproxy > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#remoteipinternalproxylist > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#remoteipproxiesheader > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#remoteiptrustedproxy > > > > > > http://httpd.apache.org/docs/current/mod/mod_remoteip.html#remoteiptrustedproxylist > > |
