Configuring a Site-to-Site IPsec VPN¶
This article covers configuring a site to site VPN link between two firewalls using IPsec. This document primarily concerns firewalls running pfSense software, but also discusses how to configure site to site links with third party IPsec-compliant devices.
The IPsec section contains example VPN Configurations that cover site to site IPsec configuration with some third party IPsec devices. If pfSense is known to work in a site to site IPsec configuration with a third party IPsec device not listed, we would appreciate a short submission containing configuration details, preferably with screenshots where applicable.
This document makes security recommendations based on current best practices. Failing to adhere to these recommendations will weaken security and may lead to a compromise of the VPN and its contents. For additional IPsec security recommendations and links to more sources, read the strongSwan Security Recommendations.
What is IPsec¶
IPsec (IP security) is a standard for providing security to IP packets via encryption and/or authentication, typically employing both. Its use in pfSense is for Virtual Private Networks (VPNs).
There are two types of IPsec VPN capabilities in pfSense, site to site and remote access (mobile).
Site to Site VPN Explained¶
A site to site VPN links two networks as if were directly connected, even if the networks are many hops apart across an untrusted circuit such as an Internet connection. To clients behind the firewalls at either end, they do not need to know a VPN is present; Clients need only attempt to contact the network on the other side. The VPN handles managing the secure connection to the other peers, including ensuring that the peer is authorized and that the traffic is encrypted between the peers.
Current best practices dictate that any circuit leaving a site must not be trusted, so a VPN is advised even over a dedicated private circuit.
While site to site VPNs are a good solution in many cases. IPsec adds processing overhead, and the Internet has far greater latency than a private network, so VPN connections are typically slower than dedicated private links (while maybe not throughput-wise, they at least have much higher latency). A point to point circuit typically has single digit latency to the other end, while a typical VPN connection will be 30-80+ ms depending on the number of hops on the Internet between the two VPN endpoints.
When deploying a VPN, if possible, stay with the same ISP for all sites. At a minimum, stay with ISPs that use the same backbone provider. Geographic proximity usually has no relation to Internet proximity. A server in the same city but on a different Internet-backbone provider could be as far away in Internet distance (hops) as a server on the other side of the continent. This difference in Internet proximity can make the difference between a VPN with 30 ms latency and one with 80+ ms latency.
Remote Access IPsec VPN¶
pfSense provides several means of remote access VPN, including IPsec, OpenVPN, and PPTP, and L2TP. Mobile IPsec functionality on pfSense has some limitations that could hinder its practicality for some deployments. pfSense supports NAT- Traversal which helps if any of the client machines are behind NAT, which is the typical case.
One good use of the pfSense IPsec client VPN capabilities is to secure all traffic sent by hosts on a wireless network or other untrusted network. This will be described later in this chapter.
Remote users can connect back to a firewall running pfSense software using a variety of different IPsec client software. See Remote Access Mobile VPN Client Compatibility
Before getting started, the following items must be complete.
The firewall must be configured and working properly for the existing local network environment.
Both locations must be using non-overlapping LAN IP subnets.
For example, if both sites are using
192.168.1.0/24on the LAN, no site to site VPN will work. This is not a limitation in pfSense, but of basic IP routing. When any host on either of the networks tries to communicate with
192.168.1.0/24, it will consider that host to be on its local LAN and the packets will never reach the firewall to be passed over the VPN connection. Similarly, if one site is using
192.168.0.0/16and one using
192.168.1.0/24, these subnets are also overlapping and a site to site VPN will not work. Keep in mind the more networks that are linked together the more important this basic fact becomes. Do not use unnecessarily large subnet masks. If the LAN is
10.0.0.0/8, but it only has 100 hosts on it, that is unnecessarily limiting the ability to add VPN networks anywhere in the
10.x.x.xspace. NAT can work around scenarios where there are conflicting subnets, though it’s preferable to avoid NAT in such circumstances and may be a requirement depending on what kind of functionality is required across the VPN.
If the firewall running pfSense is not the default gateway on the LAN where it is installed, static routes must be added to the default gateway, pointing the remote VPN subnet to an IP address on pfSense in a subnet shared between pfSense and the default gateway. Less desirable, but also functional, would be to add static/persistent routes to the client PCs who need access to the VPN.
Control of the other end of the VPN, or contact with the person who controls the other end of the VPN. If it is another firewall running pfSense, then share this document with the other administrator. If it is not, then have them consult the documentation for the IPsec device in use at the other site.
Host and application level security become more important when connecting multiple networks, depending on how much the other network is trusted. If a device on the remote network is compromised by an attacker, they could easily hop over the VPN to attack local devices without any firewall protection all traffic is allowed to pass on the VPN tunnel.
Pay attention to what is being configured! If a VPN is created to the office, and a VPN to a friend’s home network, the friend can now hop over to the company network. Or, if a friend gets infected with a worm, it could then infect local machines and continue to propagate over the VPN connection to the office. This could result in termination/firing by the company if someone else was caught on their network coming in over this VPN. Best bet here is if there is a site to site VPN into a network at work, do not connect with friends. Or use one network and firewall for accessing work and one for accessing a friend’s network.
Configuring the VPN Tunnel¶
First, log into the firewall running pfSense for the local network and click VPN > IPsec
IPsec tunnels have two components: A Phase 1 area that defines the remote peer and how the tunnel is authenticated, and one or more Phase 2 entries that define how traffic is carried across the tunnel.
If the information is incorrect in either section, the tunnel will likely fail to successfully negotiate phase 1 and/or phase 2. The trick here, as for all other parts of VPN configuration, is to make sure that both VPN servers have EXACTLY THE SAME SETTINGS for every field, with only a few exceptions to that rule: Both sides will have different a Identifier and Remote Gateway. Subnet definitions, timeouts, encryption settings, etc, all need to match.
To configure a new tunnel, a new Phase 1 must be created. Click the Add P1 button to add a new IPsec tunnel Phase 1 definition.
This is an “on / off” switch. If the tunnel should be disabled for any reason, check this option. When tunnel is needed again, uncheck it.
- Key Exchange Version
This can be IKEv1, IKEv2, or Auto.
IKEv1 is more common and widely supported, but has known issues with supporting common modern issues such as dealing with NAT or mobile clients.
An updated version of the protocol which has increased capabilities and security, as well as built-in support for mobile clients and NAT.
This option uses IKEv2 when initiating, but will accept either IKEv2 or IKEv1 when responding.
- Internet Protocol
Selects whether the tunnel will connect to an IPv4 or IPv6 remote peer. IKEv1 does not allow mixing protocols. If an IKEv1 Phase 1 will carry IPv4 traffic in Phase 2, it must also connect to an IPv4 peer on Phase 1. This limitation does not exist in IKEv2, which may carry both IPv4 and IPv6 no matter which is used on Phase 1.
This determines which part of the network will be the termination point (end point) for the IPsec tunnel. If the tunnel will be connecting to a remote server, then WAN is likely the desired setting. This can also be a virtual IP address. A gateway group can also be used for automatic failover
- Remote Gateway
This is the IP Address for the peer to which the tunnel will be established. This is most likely the WAN IP address of the remote firewall. A hostname may also be used in this field. Entering a hostname allows a tunnel to be defined between endpoints that have dynamic IP addresses.
It is a good practice to leave notes about the purpose of a tunnel. Enter a few works to describe what this VPN tunnel is used for, or about the remote end of the tunnel. This serves as a reminder for anyone managing the firewall (present or future) as to who or what will be using the tunnel.
- Authentication Method
There are two possible methods: Mutual PSK and Mutual RSA.
- Mutual PSK
Pre-Shared Key authentication. Both endpoints must create and exchange a single matching secure key to use.
- Mutual RSA
Authentication using RSA Certificates. Each peer must have a copy of the Certificate Authority used to sign the peer certificate to validate its identity and validity.
- Negotiation Mode
(IKEv1 only) This is the type of authentication security that this tunnel will use. This can be either Main or Aggressive.
More secure, but also slower and more strict. This mode is best for security, but not speed.
Less secure, but much faster and will insure that the VPN tunnel will rebuild itself quickly and probably won’t time out an application if the tunnel was down when the resource on the other end was requested.
- My Identifier
Identifies this firewall to the far side. It is best left at My IP Address and the firewall will fill it in as needed. In some cases an FQDN or similar may be entered so that the value is constant. So long as both sides agree on the identifier, it will work.
- Peer Identifier
Identifies the peer on the far side of the tunnel. It is best left at Peer IP Address and the firewall will fill it in as needed. In some cases an FQDN or similar may be entered so that the value is constant. So long as both sides agree on the identifier, it will work.
- Pre-Shared Key
(Mutual PSK authentication only) This key must be exactly the same on both VPN peers. It is case sensitive. Think of this like a “password” for the tunnel. Since this only gets entered once on each side and there is no need to remember it, it is better to make this as long and complex as possible.
This Pre-Shared Key must be as random as possible to protect the contents of the tunnel. See Securely Generating a Pre-Shared Key for details on making a secure key.
- My Certificate
(Mutual RSA authentication only) Defines the certificate which identifies this firewall. The CA which signed this certificate must be known by the peer, which may be sending them a copy of the CA certificate. If one is not shown, create or import it under System > Cert Manager on the Certificates tab.
- Peer Certificate Authority
(Mutual RSA authentication only) Defines the CA which has signed the certificate sent by the peer. This is used to validate the peer certificate. If it does not show in the list, import it under System > Cert Manager on the Certificate Authorities tab.
- Phase 1 Encryption Options
Multiple combinations of these options can be defined using the Add Algorithm button to add another line.
- Encryption Algorithm
If both sides support AES-GCM, use AES128-GCM with a 128 bit Key Length. This will combine strong encryption and hashing together and can be accelerated by AES-NI. Failing that, use AES With a Key Length of 128. If the peer does not support any of these, use the strongest available option supported by the peer.
- Hash Algorithm
Hash algorithms are used with IPsec to verify the authenticity of packet data and as a Pseudo-Random Function (PRF). When using AES-GCM, this is used solely as a PRF because AES-GCM already performs hashing internally. The best choice for use with AES-GCM is AES-XCBC. If a different type of Encryption Algorithm is in use, then use SHA256 if possible. If the peer does not support any of these, use the strongest available option supported by the peer.
- DH Key Group
We recommend not using less than DH Group 14 (2048 bit) if both sides support it. Avoid using groups 1, 2, 22, 23, and 24 as they do not provide sufficient security. As with the other options, if the suggested value is not supported by the peer, use the strongest available option.
The lifetime defines how often the connection will be rekeyed, in seconds.
28800seconds is a good balance of frequent rekeying without being too aggressive.
- Disable Rekey
Prevents the IPsec daemon from rekeying this tunnel. The far side must initiate the rekey. Leave this unchecked so that either side may initiate a rekey event.
- Margintime (Seconds)
Leave blank. Defines an alternate time frame in which a rekey attempt should be made.
- Disable Reauth
(IKEv2 Only) Skips the authentication step when performing a rekey. Faster, but less secure. Leave unchecked for stronger security.
- Responder Only
Instructs the IPsec daemon to not initiate connections, even if traffic needs to use the tunnel. The remote peer must initiate the connection. Leave unchecked so that either side may initiate as needed.
(IKEv2 Only) When enabled, allows a roaming or multi-homed peer to change IP addresses. Leave set to Disable unless this scenario is required.
- Split Connections
(IKEv2 Only) When an IKEv2 tunnel has multiple Phase 2 definitions, some peer equipment does not properly handle how the traffic selectors are generated. This is especially common in Cisco equipment. Only check this option if the remote peer cannot properly negotiate in this situation.
- NAT Traversal
(IKEv1 Only) The default Auto option will detect and use NAT Traversal when one or both peers is determined to be behind NAT. This can also be set to Force if the auto detection is not properly switching as expected.
- Dead Peer Detection
Leave enabled at the default settings. This detects when an IPsec peer has lost connectivity or otherwise is unreachable. It lets the IPsec daemon know to attempt a fresh negotiation.
Time between DPD probe attempts. The default of
- Max Failures
Number of failures before the peer is considered down. The default of
Click Save to store the Phase 1 settings. Do not click Apply Changes.
Phase 2 is what sets the parameters for traffic encryption, and defines what traffic will use the tunnel and how.
To create a new Phase 2:
Find the Phase 1 entry in the list on VPN > IPsec
Click Show Phase 2 Entries to expand the Phase 2 list
Click Add P2 to configure a new Phase 2 entry
The Phase 2 information can be filled in as follows:
An on/off switch for this Phase 2 entry only.
In almost all cases this will be a tunnel mode, such as Tunnel IPv4. With IKEv1, this will match the outer protocol of the tunnel, for example an IPv4 peer would be Tunnel IPv4. IKEv2 can have either/or (or both).
- Tunnel IPv4
A tunnel that will carry traffic between IPv4 networks.
- Tunnel IPv6
A tunnel that will carry traffic between IPv6 networks.
Encrypts all traffic between the endpoints rather than tunneling specific interesting internal traffic.
- Routed (VTI)
Routed IPsec using Virtual Tunnel Interfaces. See Routed IPsec (VTI).
- Local Network
For Tunnel IPv4/IPv6, this defines which subnet or host can be accessed from the other side of the VPN tunnel. The easiest thing to do is to set this to “LAN subnet”. This means the entire LAN will be accessible from the remote network. For Routed (VTI), this sets the local IP address and subnet mask for the
ipsecXinterface tunnel network.
The other end of the tunnel has this same field, except on the far side it is Remote Subnet. Ensure that the other end is set exactly the same. For example, if Single host is chosen in this section and the IP address of a host was entered, the other side would need to set that host in the Remote Network field.
- NAT/BINAT Translation
If the actual Local Network must be hidden from the far side, enter the settings to present to the far side (NAT+IPsec). For more details, see NAT with IPsec Phase 2 Networks.
- Remote Network
For Tunnel IPv4/IPv6 this defines which subnet or host to be accessed on the other end of the tunnel. As mentioned in the previous item, it is paramount that this is set this exactly like the other end’s Local Network section. If not, phase 2 of the VPN connection will fail and traffic will not pass from one VPN segment to the other. For Routed (VTI), this sets the remote IP address and for the
ipsecXinterface tunnel network (the peer address on the tunnel interface).
A description for this Phase 2 entry. Shows up in the IPsec status for reference.
ESP is the de facto standard for what most VPN systems use as a transport protocol. This is the recommended setting.
The firewall will automatically generate a firewall rule to allow ESP or AH to the endpoint of the VPN. If it does not, a firewall rule allowing ESP (or AH) traffic to the endpoint interface will need to be created.
- Encryption Algorithms
As before in phase 1, make sure the algorithm is set exactly as it is set on the other VPN peer. Several can be used if desired; Everything selected is available for use. That said, it is recommended to only check the one that will be used. Use AES128-GCM if available, or AES 128 otherwise.
- Hash Algorithm
As in phase 1, make sure the selected hash matches the other end. And as with the previous step, don’t add unnecessary entries. SHA256 is the preferred default, but like phase 1, some routers may only support SHA1 or MD5. When using AES-GCM, do not select any Hash Algorithm entries as AES- GCM already performs hashing.
- PFS Key Group
This works similarly to the DH group in phase 1. 14 (2048 bit) is a good setting, the default is off.
The lifetime for which the negotiated keys will be valid. One hour (
3600) is a good setting. Do not set this to too high (e.g. more than about a day:
86400) as doing so will give people more time to crack the key. Don’t be over paranoid either; there is no need to set this to 20 minutes either.
- Automatically ping host
An IP address in the remote Phase 2 network to ping to keep the tunnel alive. See What should I ping for IPsec Keep Alive for details.
Click Save, then click Apply Changes. The IPsec configuration is complete, but there are still some details to handle.
Add Firewall Rules¶
Firewall rules must exist (Firewall > Rules, IPsec tab) that govern traffic allowed to pass on the VPN tunnels. At a minimum, an allow all rule (Pass protocol any, src host any, dst host any) is needed. That said, more restrictive rules are better to enforce proper network security protocols.
If the firewall rules are too lenient then any host on the remote side will be able to directly contact any host on the local network as if they were on the same LAN.
Rules are automatically added to the WAN to allow the tunnel to connect, but if the option to disable automatic VPN rules is checked, then manual rules may be required. In that case, check the WAN rules to ensure that the traffic from the remote peer is allowed. IPsec uses UDP port 500 and 4500, and protocol ESP (or AH if set that way). If there is trouble establishing a tunnel, check the firewall logs (Status > System Logs, Firewall tab), and if blocked packets from the peer appear in the log, add appropriate rules to allow that traffic.
What if pfSense is not the main Internet Firewall?¶
In some cases there is a different firewall or router sitting between this firewall and the Internet. If this is the case it is necessary to add a port forward for ESP and UDP 500 to send the traffic to this firewall. The outside router must be able to properly handle NAT of this traffic, and some do not. A modem’s “DMZ” mode or 1:1 NAT may also help here. In this case, NAT Traversal will be needed, but the default Auto setting should be sufficient.
This may introduce routing difficulties on the internal network. More details can be found on this in the pfSense book.