Bypassing DNSBL for specific IPs
-
@horse2370 Thanks! After messing around a bit, I discovered that the order in which you put the networks/hosts makes no difference. The culprit for my problems was, indeed, the "server" prefix in the last line. Deleting it solved the issue.
-
@denx Great - as for the server: unless you want to make the modification posted by Gertjan, keep checking if make any changes to pfBlocker as it will come back.
I have a note on in my "Change Log Documents" to verify so it doesn't bite me.To help out pfdm007, do you have leading spaces in your Custom Options?
TIA
-
Finally got to do some testing....
@horse2370 : Yes if I remove evething in Unbound's custom options box, it saves. I also managed to find a workaround:
- Shutdown DNSBL/pfblockerNG, it will remove its line in Unbound's options
- Keep everything else in Unbound's options
- Remove the leading spaces
- Save (no problems!)
- Start DNSBL, it puts back its line in Unbound's options
- Remove the "server:"
- Save the options
Now I need to check if the bypassing works, but at least I got past the saving issues....
-
After all, it seems it is working now. In retrospect, I think there are still some bugs and things to iron out, and DNSBL's devs should definitely support unbound views, but at least it all works at this moment. Big thanks to you guys!
-
I'm sure this is obvious and likely just not the way you want to go, but another solution is to create a separate interface (vlan) for certain hosts and exclude this interface from using unbound (and pfBlockerNG/DNSBL) altogether. Then just allow hosts on this interface/vlan to access "respectable" external DNS servers. I did this for a gaming PC so that I didn't have to compromise my tight pfBlockerNG/DNSBL config for the sake of a single computer that wanted to talk to the world.
-
While this solution may work for some, I would prefer to still have some blocking on the additional vlan. I just really wanted the ability to have policy based blocking like the Sensei plugin offers for Opnsense. I'm currently trying it out with the $9.99 home plan, and so far it has been working as expected. It was so easy to set up as well, it just sucks having to pay 10/month :P lol Really would prefer to go back to pfsense. My goal was to find basic ad blocking on some (adult) pc's, and much stricter blocking for my kids pc's and guest devices.
I could have used other options, like pi-hole or opendns for the kids, but I want it to all go through my router and not depend on other services. Thanks for all the input so far!
-
@mdngi said in Bypassing DNSBL for specific IPs:
I'm sure this is obvious and likely just not the way you want to go, but another solution is to create a separate interface (vlan) for certain hosts and exclude this interface from using unbound (and pfBlockerNG/DNSBL) altogether
That's exactly what I wanted to do since day 1. Read my original post on this thread:
Basically I want all traffic on a specific VLAN (named DMZ with 192.168.2.0/24) totally bypass pfblocker & DNSBL in both directions (incoming and outgoing)...
Does it mean that there is a simpler/more permanent way of doing this? You need to detail how because AFAIK, all interfaces have their traffic go through the same DNS resolver (Unbound) and the distinction between filtered or not is done at the resolver level via the line :
include: /var/unbound/pfb_dnsbl.*conf
-
It's basically 4 things.
1). Configure unbound to not listen / respond to queries from clients on the excluded interface ("Network Interface" on the general tab of the DNS Resolver settings)
2). Don't configure any NAT / Port Forwarding for DNS on that interface (i.e. don't force DNS through pfSense like you probably would for the other interfaces)
3). Configure your DHCP settings for that interface with the DNS server(s) you want to use on the hosts for this zone (eg. Google, OpenDNS, etc) OR set the DNS server manually on the host.
4). Create the necessary FW rules for that interface to allow DNS out to the open internet / to the DNS servers you specified in #3.
-
If you want to bypass all pfBlocker's functionality on your DMZ subnet and you are not whitelisting. i.e. you allow all traffic from your DMZ outbound (or at least are not blocking DNS on UDP 53), then you can keep it really simple.
For DHCP hosts, just add a public DNS server(s) in the DHCP server settings for the DMZ.
For static IP Hosts, (or DHCP if you want to do it on each host) just configure a DNS server manually on each host. It doesn't matter if unbound is listening on the DMZ interface or not, if nothing gets sent to the interface ip address on UDP 53, it won't reply.Also, by still using my local unbound to resolve, I still get local host resolution, as not all my "internal" hostnames have records in my domains public Nameservers (DNS), hence my "views" configuration has
include: /var/unbound/host_entries.conf
for both dnsbl and bypass views. My certificates use hostnames as IP addresses can change, mainly my IPv6 as the prefix is allocated by Comcast dynamically. IPv4 if I decide to move stuff around :-)
That and for my own reasons (Keeps Comcast from tracking and hijacking my DNS) I use Cloudfare's DNS via DNS_over_HTTPS, which means unbound forwards all my DNS requests encrypted to Cloudfare. If you use one of there other two DNS services (1.1.1.2 or 1.1.1.3) you can also have them filter DNS requests for you based on their own reputation models.
-
Thanks for all the good info. Agreed that unbound wont reply if nothing is sent. I went down this road because I had initially NAT'd all DNS so it had to go through unbound regardless of the intended destination. I had to reverse a few things. In a nutshell, step #1 from my list is completely unnecessary (agreed). Step #2 is unnecessary if you hadn't created a port forward in the first place.
Also agreed that the downside to not going through unbound at all is the lack of local host resolution. But, on this machine it's not a big problem.
Currently using OpenDNS (or a completely different DNS for hosts going through VPN), but will play around with CloudFlare over TLS.
-
I have followed your recommended steps, but for some reasons, clients on my DMZ are not resolving (example Chrome says "DNS_PROBE_FINISHED_BAD_CONFIG")
- I have unselected DMZ interface from Unbound's Network Interfaces.
- There are no NAT or port forward applied to DMZ (as far as I know).
- The DNS servers fields under DHCP server settings for DMZ are all empty. AFAIK, when empty it will use the general DNS servers under System > General Setup which are 1.1.1.1 & 1.0.0.1 for me.
- There are 3 firewall rules on DMZ
-> Block any traffic from DMZ subnet to LAN subnet
-> Block any traffic from DMZ subnet to "This firewall"
-> Allow any traffic from DMZ to * . *Since rules are processed top-down, I'd expect the DNS requests on DMZ clients to match the last rule??? I must be missing something! If I suppress (deactivate) the second rule (block to this FW), same problem occurs.
EDIT: In order to keep the setup as straight FW as possible, and since I do not need local host resolution, I have no issue completely bypassing Unbound and using pfsense's DNS servers directly. I just want to avoid having to configure each client manually (most of the clients on DMZ are mobile devices that I will see only once....)
-
Have you tried specifying the public DNS servers in the DMZ DHCP pool config. i.e. don't leave that field blank?
I suspect that because unbound is enabled on pfSense, the DMZ DHCP server is still providing the DMZ Interface address for DNS, I don't think its aware you are not actually listening on that interface. Not able to double check at the moment.Check one of your DMZ hosts and see what DNS IP address they are using. Is it the same as the gateway?
Just a thought . . . .
-
@horse2370 : Yeah you are right.
From pfsense:
Leave blank to use the system default DNS servers: **this interface's IP** if DNS Forwarder or Resolver is enabled, otherwise the servers configured on the System / General Setup page.
Unbound being enabled, the interface's IP is forwarded as DNS server on clients, this is what I see from ipconfig on my work laptop running windows 10.
I will google how to keep unbound active for other interfaces and have the DNS servers forwarded to DMZ. I tried restarting Unbound but didnt help.
EDIT: Yeah... DNS Forward was disabled in Unbound's options..... I may slap myself behind the head. I now understand that I can revert Unbound's Advanced options to simply "server:include: /var/unbound/pfb_dnsbl.conf" and delete everything else since I no longer need views?
-
I think the help text means unbound is enabled on pfSense, not specifically the interface. i.e. globally.
I would just add the DNS servers to the DHCP server configuration for the DMZ. Keep it really simple. This is the same kind of recommendation used in Enterprise networks for Wireless Guest. That way, guest bypass corporate DNS and just get public.
You LAN clients uses unbound (with forwarding to Google) and DNSBL filters, DMZ clients use Google directly, etc.
-
@horse2370
I added the DNS servers (1.1.1.1 + 1.0.0.1) to DMZ's and the clients are now seeing the DNS servers with "ipconfig".Seems its all working fine for now! There is still a little issue with random skype call drops when connected to DMZ, I will investigate then open a separate thread if need be.
Thanks to all for the help and patience, especially you @horse2370 !
-
@horse2370 said in Bypassing DNSBL for specific IPs:
the same kind of recommendation used in Enterprise networks for Wireless Guest. That way, guest bypass corporate DNS and just get public.
Right !!
@horse2370 said in Bypassing DNSBL for specific IPs:
You LAN clients uses unbound (with forwarding to Google)
Euh .... Corporate clients networks forwarding to Google ??
Wrong
At least in Europe, that's not a recommendation thing, one could get fired for that. Maybe in the States other rules or reasons exist, I can't tell.
For me, "8.8.8.8" is at maximum a soHO thing.
(I really guess because our society fabricated a lot of people that are trained to "have to enter a DNS" because our ISP's trained us to do so in the past. They had their reasons, but these do not exist any more).But if someone can tell me ones and for all what the benefits are, I'll adopt "8.8.8.8" myself, promised.
Btw : don't get me wrong : I'm using gmail (just perfect for automated notification systems).
I do like their search engine a lot - it works for me
I'm not against Google as a company (how could I). -
Great! Your Welcome
Thanks for clarifying in case others also thought I recommended Enterprise customers forward to Google from their internal DNS. What I thought I wrote was ONLY referring to Guest network recommendation.
My last sentence in that post was specific to pftdm007's environment, although it should have started with "Your" and not "You" :-)
To clarify - The general recommendation is to keep guest traffic isolated from internal resources, i.e use the internal DHCP servers on the network infrastructure, such as the WLAN controllers as typically that is where the capture portal is anyway, and have them use an ISP or Public DNS (i.e. non-enterprise), instead of letting the guest traffic roam around the datacenter/datacentre servers before being routed (you can pronounce "routed" correctly :-) ) out the Internet gateway.
For the Enterprise's own DNS, it should follow standard best practices, which as you correctly point out would not typically include forwarding all your DNS traffic to Google's Public server to data mine.
As always the answer is never quite that simple and it always depends. Some smaller companies may choose to utilize a Public/Cloud DNS service as part of their security posture, but that would be an authoritative name service. Hence, one of the reasons Cisco acquired openDNS, or you could even use Google Cloud DNS among many others.Like you, I use Google, however when it comes to privacy. . . . . . not so much. Yes I'm in the states, but originally from the UK and deal with customers World Wide and am exposed to all kinds of regulations. Most good, some just plain political and a PIA. The amount of time spent on training for compliance is getting beyond a joke. Soon I won't have time to do actual work that might contravene a regulation.
My home DNS used to resolve from root severs and I have a managed provider for my personal domains for simplicity and resiliance. I certainly avoid forwarding to my ISP's DNS and even using root servers when my ISP started hijacking DNS on UDP 53.
Quick Google for a supporting article provides some details Comcast hijacking DNS
I switched to encrypted DNS on community supported DNS servers, until Cloudfare rolled out their free public service for people like me.I could see one benefit of using Google for your DNS, they would mine those requests and know what you needed to buy before you did
As with anything, there are exceptions to every rule, your mileage may vary and it will always depend!
Stay safe
-
I just realized... DNSBL can be bypassed by passing different DNS servers to my DMZ clients (so they do not go thru Unbound), but what about pfBlockerNG?
WAN is selected in Inbound, and LAN is selected for outbound, but not DMZ (so not to have pfblocker block traffic on DMZ).
In pfsense's FW logs, I see entries showing blocked traffic on WAN that is going to DMZ. This is my understanding: DMZ client sends traffic to the internet, nothing is blocked (pfblocker & DNSBL are not running on DMZ). Request gets sent to the internet, the response comes back, it is blocked by pfblocker on WAN.
I had Snort running on LAN and WAN, but not on DMZ. It kept blocking incoming traffic on WAN that was coming back to DMZ. User @bmeeks kindly explained that it was unnecessary to have Snort run on WAN. The logic is that all unsolicited inbound traffic is blocked by FW rules so only stateful responses are allowed, and since the initial traffic had gone thru Snort on LAN, it is considered "safe". Sorry for the gross over-simplification......
Is it the same logic for pfblockerNG??? How do you guys run these packages? I'm slowly realizing that I have made several mistakes in setting up my pfsense box....
-
Correct, if any client can bypass unbound on the pfSense, it will not benefit from the pfBlockNG functionality. Hence on my pfSense deployment at home, I have a FW rule on the LAN, DMZ and VPN interfaces that only allow UDP 53 (DNS) to pass if the destination is the local interface address. I have another rule that denies and logs any other requests going to other DNS addresses which I can see in the logs and my splunk setup. (Allow/Permit rule is before the deny and log rule) This shows Amazon and Google assistants trying their hardest along with my Roku's, same is true for them trying to track and report usage statistics.
Regarding you other question about pfBlockerNG, I'm going to over simplify as that is the other function of pfBlockerNG. (i.e. it does DNS filtering and the other function is IP address filtering) If you have setup the feeds, you should see "auto rule" in the Firewall Rules admin page.
Again on my setup, I am blocking Banned Hosts, Emerging threat hosts inbound on the WAN and outbound on all other interfaces. In addition I only allow inbound connection from specific countries that have family or other services that need access to the DMZ to pass. (i.e. Mail Relay and Plex) If I travel outside these normal countries I do have to modify this rule so I can still VPN and access my DMZ services. This is all done via the GeoIP admin pages.
Your other question about inbound traffic being blocked to the DMZ, I will again over simplify, but the basic premise of the FW functionality is, any session that is initiated from a "trusted" interface (LAN, DMZ) will be allowed and the inbound traffic for that session will also be allowed back through the WAN interface. For TCP, this is tracked my SYN's and FIN's for when to allow that port and when to close it. UDP relies on the same ports being used (src/dst flipped) for the return traffic and typically timers. Therefore if you are seeing blocked traffic ( and assuming you have a fairly out of the box config) those are typically from session you have not initiated or unsolicited There are many exception to this oversimplification, mainly with Video and voice protocols.
In summary, my setup uses pfBlockerNG's DNS functionality to block ads, tracking, malware and phishing sites. My logs show the PiHole feed block 95% of them. Side effect is faster page loading and a significant drop in downloaded traffic which is good as I have a 1TB cap per month.
pfBlockerNG's IP functionality keeps bad hosts and restricts countries that should not be connecting, out of my DMZ. And stops inside hosts from accidentally connecting to bad hosts. The GeoIP/Country filter does not apply outbound obviously. The Firehol feed does most of the heavy lifting based on the reports and splunk dashboards.For me, this strikes a balance between family usability of the Internet and keeping it reasonably safe and secure.
Side note, I do have Snort installed, however it is is IDS (monitor) mode with a couple of http SID's suppressed as they caused many false positives, end result, I very few alerts and check for unusual alerts every few days.
Snort is enabled on my WAN, along with a number of FW rules that log, this is only so I can see what kind of attacks are occurring for research purposes and track the percentage of encrypted IPv4 and IPv6 traffic. This is captured in dashboards that I have configured in splunk and I use for work.
Encrypted traffic continues to rise as does the split between IPv4 and IPv6. The later being significantly offset to IPv4 when the family is streaming more as the Roku's are v4 only.Hope that helps . . . . .
-
I still wonder if I am chasing a ghost or something real... Here's an actual real life example:
My work laptop is connected to DMZ. Connection straight to the outside, no DNSBL running on DMZ, no pfblocker running on DMZ, no Snort, nothing. Straight out with the exception of a few FW rules blocking anything nasty (connecting to pfsense, communicating with other subnets, etc).
This laptop is totally dysfunctional. Intermittently, web pages dont load, stuff breaks and crashes, proxy app keeps re-creating new connections to its mother ship, etc. My employer's IT support are blaming my "home network". This laptop runs a proxy service that connects to "gateway.zscalertwo.com" on port 18000.
In pfsense's FW logs, I see hundreds and hundreds of entries showing blocked traffic on DMZ. All of these have "Default deny rule IPv4" as description. That doesnt say much....
Since the FW rules on DMZ are very basic and pretty much allow anything, I started searching for an explanation somewhere else... I found in pfBlockerNG > Reports > IP Block Stats:
Clicking on the filter icon:
So I'm going back to my original thought. To me pfblockerNG interferes with traffic on DMZ. How can I disable this without whitelisting stuff???