" deny" rule getting bypassed

Rick_H · ‎08-08-2014

I have an Action:DENY policy setup at the top of the ruleset for inbound traffic from WAN1 to DMZ. The source address for this policy is a group which consists of some geography-based networks and " bad addresses" that come up frequently in the IPS. However, this policy doesn' t ever seem to get matched and hosts I have explicitly added to the blacklist group--or even as their own entry--are matched against policies further down in the chain. I know this has to be something easy, but after staring at it for a couple of days I' m just not seeing it. Hopefully one of you fellows can see what I' m missing. This is on a FortiGate 80C running FortiOS 5.0.7 (build 3608). I' ve attached a screenshot of the rule set, but the relevant bits of code are here:

 config firewall address
     edit " all" 
     next
     edit " outside.blacklist.121.139.0.0" 
         set subnet 121.139.0.0 255.255.0.0
     next
     edit " outside.blacklist.96.8.126.97" 
         set subnet 96.8.126.0 255.255.255.0
     next
     edit " outside.blacklist.108.61.91.211" 
         set subnet 108.61.91.0 255.255.255.0
     next
 end
 config firewall addrgrp
     edit " outside.blacklist.All" 
         set member " outside.blacklist.96.8.126.97"  " outside.blacklist.108.61.91.211"  " outside.blacklist.121.139.0.0"  [geography-based lists]
     next
 end
 
 config firewall policy
     edit 53
         set srcintf " wan1" 
         set dstintf " dmz" 
         set srcaddr " outside.blacklist.All" 
         set dstaddr " all" 
         set schedule " always" 
         set service " ALL" 
         set logtraffic all
     next
     edit 26
         set srcintf " wan1" 
         set dstintf " dmz" 
         set srcaddr " all" 
         set dstaddr " nat.inbound.rounding" 
         set action accept
         set schedule " always" 
         set service " HTTP"  " HTTPS" 
         set utm-status enable
         set av-profile " MainAV" 
         set ips-sensor " protect_http_server" 
         set profile-protocol-options " default" 
     next
     edit 37
         set srcintf " wan1" 
         set dstintf " dmz" 
         set srcaddr " all" 
         set dstaddr " nat.inbound.smtp" 
         set action accept
         set schedule " always" 
         set service " SMTP" 
         set utm-status enable
         set av-profile " MainAV" 
         set spamfilter-profile " MSPS Email Filter" 
         set ips-sensor " protect_email_server" 
         set profile-protocol-options " default" 
     next
     edit 42
         set srcintf " wan1" 
         set dstintf " dmz" 
         set srcaddr " all" 
         set dstaddr " nat.inbound.owa" 
         set action accept
         set schedule " always" 
         set service " HTTPS"  " HTTP" 
         set utm-status enable
         set av-profile " MainAV" 
         set ips-sensor " protect_http_server" 
         set profile-protocol-options " default" 
     next
     edit 49
         set srcintf " wan1" 
         set dstintf " dmz" 
         set srcaddr " all" 
         set dstaddr " nat.inbound.rdsgate" 
         set action accept
         set schedule " always" 
         set service " HTTPS"  " RDS_external_UDP" 
         set utm-status enable
         set av-profile " MainAV" 
         set ips-sensor " protect_http_server" 
         set profile-protocol-options " default" 
     next
     edit 51
         set srcintf " wan1" 
         set dstintf " dmz" 
         set srcaddr " all" 
         set dstaddr " nat.inbound.remote" 
         set action accept
         set schedule " always" 
         set service " HTTPS" 
         set utm-status enable
         set av-profile " MainAV" 
         set ips-sensor " protect_http_server" 
         set profile-protocol-options " default" 
     next
 end

These policies have been in place for sometime, yet this morning I have had the same IP address showing up in my IPS logs quite a bit:

Message meets Alert condition The following intrusion was observed: " PHP.CGI.Argument.Injection" . date=2014-08-08 time=10:18:04 devname=fw01 devid=FGT80C3911608826 logid=0419016384 type=ips subtype=signature level=alert severity=high srcip=96.8.126.97 dstip=[DMZ address of host in policy 42 above] srcintf=" wan1" dstintf=" dmz" policyid=42 identidx=0 sessionid=13078459 status=dropped proto=6 service=http count=1 attackname=" PHP.CGI.Argument.Injection" srcport=33113 dstport=80 attackid=31752 sensor=" protect_http_server" ref=" http://www.fortinet.com/ids/VID31752" incidentserialno=513620505 msg=" web_server: PHP.CGI.Argument.Injection,"

The source IP in this message is in my blacklist and I' ve tried specifying both it as a host and a /24 which included the address. Yet, inbound traffic from this host still seems to whiff past the deny policy and match against the next policy in the list instead. Any idea what I' m missing here? Any help at all is appreciated. Cheers, Rick

Dave_Hall · ‎04-09-2015

If you are trying to block countries by geo, I would do it via local-in-policy., e.g.

config system global
    set gui-local-in-policy enable
end
config firewall address
    edit "block-country-netherlands"
        set associated-interface "wan1"
        set type geography
        set country "NL"
    next
    edit "block-country-brazil"
        set associated-interface "wan1"
        set type geography
        set country "BR"
    next
    edit "block-country-lithuania"
        set associated-interface "wan1"
        set type geography
        set country "LT"
    next
    edit "block-country-sweden"
        set associated-interface "wan1"
        set type geography
        set country "SE"
    next
    edit "block-country-china"
        set associated-interface "wan1"
        set type geography
        set country "CN"
    next
    edit "block-country-russia"
        set associated-interface "wan1"
        set type geography
        set country "RU"
    next
end
config firewall addrgrp
    edit "block-countries-group"
        set member "block-country-brazil" "block-country-china" "block-country-lithuania" "block-country-netherlands" "block-country-russia" "block-country-sweden"
    next
end
config firewall local-in-policy
    edit 0
        set intf "wan1"
        set srcaddr "block-countries-group"
        set dstaddr "all"
        set service "ALL"
        set schedule "always"
    next
end

NSE4/FMG-VM64/FortiAnalyzer-VM/6.0 (FWF30E/FW92D/FGT200D/FGT101E/FGT81E)/ FAP220B/221C

View solution in original post

trauthor · ‎06-18-2015

There are multiple posts in this forum related to VIP policy compromising security. I have opened cases with FortiNet both for this issue and an additional issue. I have also alerted appropriate parties. To FortiNet's credit, they are working quickly to address this. I have not yet seen an official public statement from FortiNet. Please refer to the below correspondence to see if it pertains to your situation. Thank you.

(O.P.: not sure if it matters at this point, but I don't see a "set action deny" for rule 53 in the copy you took from the CLI.)

==========

Fw: FortiGate Security "Loophole" and Severe Bug

Two issues were discovered during FortiGate firewall product tests, the first a documentation issue which FortiNet has confirmed affects FortiOS 5.0.x and 5.2.x and the second a bug which affects any FortiGate "D" series in combination with FortiOS 5.0.10 (the FIPS 140 version; it is unknown whether other combinations of FortiOS and FortiGate are affected.)

1) FortiGate Deny All rules do not deny all traffic. What is documented: "VIP rules" take precedence over "regular rules." However, until two days ago (6/15/2015) after it was recently brought to their attention this was mentioned only briefly in a technical note and not in any of their standard documentation (the FortiOS handbook, admin guide, etc.) It remains inexplicit that "VIP rules" also take precedence over "Deny All" rules.

- Here's the link to the technical note (taken from the support case): http://kb.fortinet.com/kb/microsites/search.do?cmd=displayKC&docType=kc&externalId=FD33338

- Here's the link to the updated handbook, published 6/15/2015 (see page 956, "Exception to policy order (VIPs):") http://docs.fortinet.com/d/fortigate-fortios-handbook-the-complete-guide

The scenario documented in the stem support case is given below. Rules appear in the same order they would after issuing the commands "config firewall policy" and "get."

=====

config firewall policy

edit 1

set srcintf "wan1"

set srcaddr "outside_blacklist"

set dstintf "dmz"

set dstaddr "all"

set action deny

set schedule "always"

set service "ALL"

set logtraffic all

set srcintf "wan1"

set srcaddr "all"

set dstintf "dmz"

set dstaddr "nat_inside"

set action accept

set schedule "always"

set service "ALL"

set logtraffic all

=====

In the above example, "outside_blacklist" is a group of outside addresses and "nat_inside" is a VIP on the firewall. As depicted, traffic will not follow the usual order of precedence. Any traffic destined for "nat_inside" will ignore the Deny All in Rule 1 but match the Accept in Rule 2. Below is the fix suggested by a FortiNet support engineer.

=====

config firewall policy

edit 1

set srcintf "wan1"

set srcaddr "outside_blacklist"

set dstintf "dmz"

set dstaddr "nat_inside"

set action deny

set schedule "always"

set service "ALL"

set logtraffic all

set srcintf "wan1"

set srcaddr "all"

set dstintf "dmz"

set dstaddr "nat_inside"

set action accept

set schedule "always"

set service "ALL"

set logtraffic all

=====

Shown above, the change replaces the Deny All in Rule 1 with Deny "nat_inside." This in effect restores the order of precedence. (For an explanation, see the new section "Exception to policy order (VIPs)" in the handbook. For the alternative fix, "match-vip," see the technical note.) To summarize, FortiGate "VIP rules" are matched before "regular rules." Most significantly, "VIP rules" are matched before "Deny All" rules.

The potential for a security "loophole" as described in the above case is compounded by:

a) a reasonable (until two days ago) expectation that a Deny or Deny All policy always denies blacklisted traffic in order of precedence; and

b) typical industry practice that only denied traffic gets logged.

Given these factors, a configuration error which would allow blacklisted traffic to enter a network undetected via an Accept rule is not unlikely. [...]

2) The secondary unit in a FortiGate "D" series active/passive cluster bricks (i.e., fails closed and must be re-imaged) after the first couple FIPS self-tests under certain conditions, two of them being: when it can't contact the master; when it is given the master's configuration file. Anyone with "D" series units will not be able to maintain an active/passive cluster in FIPS-CC mode. An "emergency" code fix of FortiOS 5.0.10 is underway which will be released as 5.0.13. We have been assured by FortiNet this will not affect the FIPS 140 certification of FortiOS.

View solution in original post

trauthor · ‎06-18-2015

"Fantastic summary trauthor. From your post, I get the impression that Fortinet will be re-working the flow of traffic such that a deny would somehow take precedence, is this indeed the case?"

The only response of which I was made aware is the documentation update just published on 6/15, which explains that rules which contain a VIP do not follow the usual order of precedence. Originally, they were going to relegate the update to the cookbook, until someone there impressed upon them that there were security implications. I was not given the impression there was any other action they were going to take (other than correcting the documentation.) I should probably amend my statement about FortiNet working quickly to correct this to say that they updated the standard documentation on 6/15 and are working on the second issue (FIPS cluster failures). Personally, I'd still like the document to be a little clearer (e.g., regarding Deny All rules and "match-vip"), but mostly I wanted to get the issue some visibility because of the serious risks to security.

trauthor · ‎06-26-2015

FYI

Congrats to the O.P. who started this thread and all those who responded. Please see a snippet of the official response from cert.org (the place US-CERT vulnerability requests go) pertaining to the rule precedence / match-vip documentation gap. Please note also that FortiNet has corrected the documentation, which was re-published on 6/15.

-----BEGIN PGP SIGNED MESSAGE----- [...]

Hello [redacted],

While we [...] do not intend to publish a vulnerability note on this issue, we have taken the following actions:

1) We have made US-CERT aware of the issue to notify potentially affected US departments and agencies.

2) We have sent Fortinet our recommendation that in addition to the documentation updates, they should notify their customers.

In response, Fortinet said that they will "take the necessary steps to document the issue more clearly."

Regards,

[redacted] Vulnerability Analysis Team ======================================================================

CERT Coordination Center [...]

djwilliams · ‎06-27-2016

I was hoping for some clarification on one point. The implicit Deny "ALL" at the bottom is always matched regardless of any policy or policy order above it correct? If it makes it to the bottom of the policy list without matching something it will match the implicit. Is it safe to say then that "match-vip" is disabled by default on all policies EXCEPT the implicit deny?

Network Engineer

Marcelo_Clementeen · ‎11-10-2016

Thanks for the feedback.
My question is:
I need the port2 (WAN) of the fortigate to make available the public 5 ips, because I can only ping the ip when enabled in the interface configuration of Fortigate. 
Being from within the fortigate I can ping all public IPs, but only outside the gateway (200.222.166.1). 
Thank you for your help

emnoc · ‎11-10-2016

This is not a bug IMHO but lack of understand of fw policies and with regards to vip and non-vip fwpolicies

"

The FortiGate unit matches Virtual IP firewall policies differently from regular firewall policies. If there is a VIP firewall policy below a "regular" DENY firewall policy, the VIP traffic will still be able to go through."

http://kb.fortinet.com/kb/documentLink.do?externalID=FD33338

match-vip is need for any vip when you have "specific" deny as a action and vips. I don't think this has changed from model types or fortiOS versions.

Ken

PCNSE

NSE

StrongSwan

" deny" rule getting bypassed

Nominate a Forum Post for Knowledge Article Creation