Internet Protocol Security (IPsec) is a set of protocols defined by the Internet Engineering Task Force (IETF) to secure packet exchange over unprotected IP/IPv6 networks such as Internet.
IpSec protocol suite can be divided in following groups:
- Authentication Header (AH) RFC 4302
- Encapsulating Security Payload (ESP) RFC 4303
- Internet Key Exchange (IKE) protocols. Dynamically generates and distributes cryptographic keys for AH and ESP.
Authentication Header (AH)
AH is a protocol that provides authentication of either all or part of the contents of a datagram through the addition of a header that is calculated based on the values in the datagram. What parts of the datagram are used for the calculation, and the placement of the header, depends whether tunnel or transport mode is used.
The presence of the AH header allows to verify the integrity of the message, but doesn't encrypt it. Thus, AH provides authentication but not privacy (Another protocol ESP is used to provide encryption).
RouterOS supports the following authentication algorithms for AH:
In transport mode AH header is inserted after IP header. IP data and header is used to calculate authentication value. IP fields that might change during transit, like TTL and hop count, are set to zero values before authentication.
In tunnel mode original IP packet is encapsulated within a new IP packet. All of the original IP packet is authenticated.
Encapsulating Security Payload
Encapsulating Security Payload (ESP) uses shared key encryption to provide data privacy. ESP also supports its own authentication scheme like that used in AH, or can be used in conjunction with AH.
ESP packages its fields in a very different way than AH. Instead of having just a header, it divides its fields into three components:
- ESP Header - Comes before the encrypted data and its placement depends on whether ESP is used in transport mode or tunnel mode.
- ESP Trailer - This section is placed after the encrypted data. It contains padding that is used to align the encrypted data.
- ESP Authentication Data - This field contains an Integrity Check Value (ICV), computed in a manner similar to how the AH protocol works, for when ESP's optional authentication feature is used.
In transport mode ESP header is inserted after original IP header. ESP trailer and authentication value is added to the end of the packet. In this mode only IP payload is encrypted and authenticated, IP header is not secured.
In tunnel mode original IP packet is encapsulated within a new IP packet thus securing IP payload and IP header.
RouterOS ESP supports various encryption and authentication algorithms.
- DES - 56-bit DES-CBC encryption algorithm;
- 3DES - 168-bit DES encryption algorithm;
- AES - 128, 192 and 256-bit key AES-CBC encryption algorithm;
- Blowfish - added since v4.5
- Twofish - added since v4.5
- Camellia - 128, 192 and 256-bit key Camellia encryption algorithm added since v4.5
Hardware encryption allows to do faster encryption process by using built-in encryption engine inside CPU. AES is the only algorithm that will be accelerated in hardware.
List of RouterBoards with enabled hardware support:
For comparison RB1000 with enabled HW support can forward up to 550Mbps encrypted traffic. When HW support is disabled it can forward only 150Mbps encrypted traffic in AES-128 mode.
Some configuration advices on how to get maximum ipsec throughput on multicore RB1100AHx2:
- Avoid using ether12 and ethet13. Since these prots are pci-x they will be slowest ones.
- Fastest forwarding is from switch chip ports (ether1-ether10) to ether11 (directly connected to CPU) and vice versa.
- Set hardware queue on all interfaces
/queue interface set [find] set queue=only-hardware-queue
- Disable RPS:
/system resource irq rps disable [find]
- Assign one CPU core to ether11 and other CPU core to everything else. Forwarding over ether11 requires more CPU that is why we are giving one core only for that interface (in IRQ setting ether11 is listed as ether12 tx,rx and error).
/system resource irq set [find] cpu=1 set [find users="eth12 tx"] cpu=0 set [find users="eth12 rx"] cpu=0 set [find users="eth12 error"] cpu=0
- disable connection tracking
With all above recommendations it is possible to forward 820Mbps (1470byte packets two streams).
With enabled connection tracking 700Mbps (1470 byte packets two streams).
Internet Key Exchange Protocol
The Internet Key Exchange (IKE) is a protocol that provides authenticated keying material for Internet Security Association and Key Management Protocol (ISAKMP) framework. There are other key exchange schemes that work with ISAKMP, but IKE is the most widely used one. Together they provide means for authentication of hosts and automatic management of security associations (SA).
Most of the time IKE daemon is doing nothing. There are two possible situations when it is activated:
There is some traffic caught by a policy rule which needs to become encrypted or authenticated, but the policy doesn't have any SAs. The policy notifies IKE daemon about that, and IKE daemon initiates connection to remote host. IKE daemon responds to remote connection. In both cases, peers establish connection and execute 2 phases:
- Phase 1 - The peers agree upon algorithms they will use in the following IKE messages and authenticate. The keying material used to derive keys for all SAs and to protect following ISAKMP exchanges between hosts is generated also.
- Phase 2 - The peers establish one or more SAs that will be used by IPsec to encrypt data. All SAs established by IKE daemon will have lifetime values (either limiting time, after which SA will become invalid, or amount of data that can be encrypted by this SA, or both).
There are two lifetime values - soft and hard. When SA reaches it's soft lifetime treshold, the IKE daemon receives a notice and starts another phase 2 exchange to replace this SA with fresh one. If SA reaches hard lifetime, it is discarded.
IKE can optionally provide a Perfect Forward Secrecy (PFS), which is a property of key exchanges, that, in turn, means for IKE that compromising the long term phase 1 key will not allow to easily gain access to all IPsec data that is protected by SAs established through this phase 1. It means an additional keying material is generated for each phase 2.
Generation of keying material is computationally very expensive. Exempli gratia, the use of modp8192 group can take several seconds even on very fast computer. It usually takes place once per phase 1 exchange, which happens only once between any host pair and then is kept for long time. PFS adds this expensive operation also to each phase 2 exchange.
Diffie-Hellman (DH) key exchange protocol allows two parties without any initial shared secret to create one securely. The following Modular Exponential (MODP) and Elliptic Curve (EC2N) Diffie-Hellman (also known as "Oakley") Groups are supported:
|Group 1||768 bit MODP group||RFC 2409|
|Group 2||1024 bits MODP group||RFC 2409|
|Group 3||EC2N group on GP(2^155)||RFC 2409|
|Group 4||EC2N group on GP(2^185)||RFC 2409|
|Group 5||1536 bits MODP group||RFC 3526|
To avoid problems with IKE packets hit some SPD rule and require to encrypt it with not yet established SA (that this packet perhaps is trying to establish), locally originated packets with UDP source port 500 are not processed with SPD. The same way packets with UDP destination port 500 that are to be delivered locally are not processed in incoming policy check.
To get IPsec to work with automatic keying using IKE-ISAKMP you will have to configure policy, peer and proposal (optional) entries.
/ip ipsec peer
Peer configuration settings are used to establish connections between IKE daemons (phase 1 configuration). This connection then will be used to negotiate keys and algorithms for SAs.
|address (IP/IPv6 Prefix; Default: 0.0.0.0/0)||If remote peer's address matches this prefix, then the peer configuration is used in authentication and establishment of Phase 1. If several peer's addresses match several configuration entries, the most specific one (i.e. the one with largest netmask) will be used.|
|auth-method (pre-shared-key | rsa-signature; Default: pre-shared-key)||Authentication method:
|certificate (string; Default: )||Name of a certificate listed in certificate table (signing packets; the certificate must have private key). Applicable if RSA signature authentication method (auth-method=rsa-signature) is used.|
|comment (string; Default: )||Short description of the peer.|
|dh-group (ec2n155 | ec2n185 | modp1024 | modp1536 | modp2048 | modp3072 | modp4096 | modp6144 | modp768; Default: modp1024)||Diffie-Hellman group (cipher strength)|
|disabled (yes | no; Default: no)||Whether peer is used to match remote peer's prefix.|
|dpd-interval (time | disable-dpd; Default: 2m)||Dead peer detection interval. If set to disable-dpd, dead peer detection will not be used.|
|dpd-maximum-failures (integer: 1..100; Default: 5)||Maximum count of failures until peer is considered to be dead. Applicable if DPD is enabled.|
|enc-algorithm (3des | aes-128 | aes-192 | aes-256 | blowfish | camellia-128 | camellia-192 | camellia-256 | des; Default: 3des)||Encryption algorithm.|
|exchange-mode (aggressive | base | main | main-l2tp; Default: main)||Different ISAKMP phase 1 exchange modes according to RFC 2408. Do not use other modes then main unless you know what you are doing. main-l2tp mode relaxes rfc2409 section 5.4, to allow pre-shared-key authentication in main mode.|
|generate-policy (yes | no; Default: no)||Allow this peer to establish SA for non-existing policies. Such policies are created dynamically for the lifetime of SA. Automatic policies allows, for example, to create IPsec secured L2TP tunnels, or any other setup where remote peer's IP address is not known at the configuration time.|
|hash-algorithm (md5 | sha1; Default: md5)||Hashing algorithm. SHA (Secure Hash Algorithm) is stronger, but slower.|
|key (string; Default: )||Name of the key from key menu. Applicable if auth-method=rsa-key.|
|lifebytes (Integer: 0..4294967295; Default: 0)||Phase 1 lifetime: specifies how much bytes can be transferred before SA is discarded. If set to 0, SA will not be discarded due to byte count excess.|
|lifetime (time; Default: 1d)||Phase 1 lifetime: specifies how long the SA will be valid.|
|my-id-user-fqdn (string; Default: )||By default IP address is used as ID. This parameter replaces ID with specified value. Can be used, for example, in cases if DNS name as ID is required.|
|nat-traversal (yes | no; Default: no)||Use Linux NAT-T mechanism to solve IPsec incompatibility with NAT routers inbetween IPsec peers. This can only be used with ESP protocol (AH is not supported by design, as it signs the complete packet, including IP header, which is changed by NAT, rendering AH signature invalid). The method encapsulates IPsec ESP traffic into UDP streams in order to overcome some minor issues that made ESP incompatible with NAT.|
|port (integer:0..65535; Default: 500)||Communication port used for ipsec traffic.|
|proposal-check (claim | exact | obey | strict; Default: obey)||Phase 2 lifetime check logic:
|remote-certificate (string; Default: )||Name of a certificate (listed in certificate table) for authenticating the remote side (validating packets; no private key required). Applicable if RSA signature authentication method is used|
|secret (string; Default: )||Secret string (in case pre-shared key authentication is used). If it starts with '0x', it is parsed as a hexadecimal value|
|send-initial-contact (yes | no; Default: yes)||Specifies whether to send initial IKE information or wait for remote side.|
/ip ipsec key
This submenu list all imported public/private keys, that can be used for peer authentication. Submenu also has several commands to work with keys.
For example print below shows two imported 1024-bit keys, one public and one private.
[admin@PoETik] /ip ipsec key> print Flags: P - private-key, R - rsa # NAME KEY-SIZE 0 PR priv 1024-bit 1 R pub 1024-bit
|export-pub-key (file-name; key)||Export public key to file from one of existing private keys.|
|generate-key (key-size; name)||Generate private key. Takes two parameters, name of newly generated key and key size 1024,2048 and 4096.|
|import (file-name; name)||Import key from file.|
/ip ipsec policy
Policy table is used to determine whether security settings should be applied to a packet.
|action (discard | encrypt | none; Default: encrypt)||Specifies what to do with packet matched by the policy.
|comment (string; Default: )||Short description of the policy|
|disabled (yes | no; Default: no)||Whether policy is used to match packets.|
|dst-address (IP/IPv6 prefix; Default: 0.0.0.0/32)||Destination address to be matched in packets.|
|dst-port (integer:0..65535 | any; Default: any)||Destination port to be matched in packets. If set to any all ports will be matched|
|ipsec-protocols (ah | esp; Default: esp)||Specifies what combination of Authentication Header and Encapsulating Security Payload protocols you want to apply to matched traffic|
|level (require | unique | use; Default: require)||Specifies what to do if some of the SAs for this policy cannot be found:
|manual-sa (string | none; Default: none)||Name of the manual SA template|
|priority (integer:-2147483646..2147483647; Default: 0)||Policy ordering classificator (signed integer). Larger number means higher priority.|
|proposal (string; Default: default)||Name of the proposal template that will be sent by IKE daemon to establish SAs for this policy.|
|protocol (all | egp | ggp| icmp | igmp | ...; Default: all)||IP packet protocol to match.|
|sa-dst-address (ip/ipv6 address; Default: ::)||SA destination IP/IPv6 address (remote peer).|
|sa-src-address (ip/ipv6 address; Default: ::)||SA source IP/IPv6 address (local peer).|
|src-address (ip/ipv6 prefix; Default: 0.0.0.0/32)||Source IP prefix|
|src-port (any | integer:0..65535; Default: any)||Source Port of the packet|
|tunnel (yes | no; Default: no)||Specifies whether to use tunnel mode|
/ip ipsec policy print stats will show current status of the policy. Additional read-only parameters will be printed.
|in-accepted (integer)||How many incoming packets were passed by the policy without an attempt to decrypt.|
|in-dropped (integer)||How many incoming packets were dropped by the policy without an attempt to decrypt|
|in-transformed (integer)||How many incoming packets were decrypted (ESP) and/or verified (AH) by the policy|
|out-accepted (integer)||How many outgoing packets were passed by the policy without an attempt to encrypt.|
|out-dropped (integer)||How many outgoing packets were dropped by the policy without an attempt to encrypt.|
|out-transformed (integer)||How many outgoing packets were encrypted (ESP) and/or verified (AH) by the policy.|
|ph2-state (expired | no-phase2 | established)||Indication of the progress of key establishing.|
It is possible to dump policies installed into the kernel for debugging purposes with command:
/ip ipsec policy dump-kernel-policies
After executing this command check the logs to see the result, there should be three policies in the kernel: forward, in and out.
[admin@test-host] >/log print 07:28:34 ipsec,debug,packet policy ipsec fwd: 10.5.101.9 - 10.5.101.13 07:28:34 ipsec,debug,packet policy ipsec in: 10.5.101.9 - 10.5.101.13 07:28:34 ipsec,debug,packet policy ipsec out: 10.5.101.13 - 10.5.101.9
/ip ipsec proposal
Proposal information that will be sent by IKE daemon to establish SAs for this policy. Configured proposals are set in policy configuration.
|auth-algorithms (md5|sha1|null; Default: sha1)||Allowed algorithms for authorization. sha1 is stronger, but slower algorithm.|
|comment (string; Default: )||Short description of an item.|
|disabled (yes | no; Default: no)||Whether item is disabled.|
|enc-algorithms (null|des|3des|aes-128|aes-192|aes-256|blowfish|camellia-128|camellia-192|camellia-256|twofish; Default: 3des)||Allowed algorithms and key lengths to use for SAs.|
|lifetime (time; Default: 30m)||How long to use SA before throwing it out.|
|name (string; Default: )||Name of the proposal template, that will be identified in other parts of ipsec configuration.|
|pfs-group (ec2n155 | ec2n185 | modp1024 | modp1536 | modp2048 | modp3072 | modp4096 | modp6144 | modp768 | none; Default: modp1024)||Diffie-Helman group used for Perfect Forward Secrecy.|
/ip ipsec manual-sa
Menu is used to configure SAs manually. Created SA template then can be used in policy configuration.
in,out = md5|null|sha1; Default: null)
|Authentication Header encryption algorithm.|
|ah-key (string/string; Default: )||Incoming-authentication-key/outgoing-authentication-key|
|ah-spi (0x100..FFFFFFFF/0x100..FFFFFFFF; Default: 0x100)||Incoming-SA-SPI/outgoing-SA-SPI|
|disabled (yes | no; Default: no)||Defines whether item is ignored or used|
in,out = md5|null|sha1; Default: null)
|Encapsulating Security Payload authentication encryption algorithm|
|esp-auth-key (string/string; Default: )||Incoming-authentication-key/outgoing -authentication-key|
in,out = 3des | aes-128 | aes-192 | aes-256 | des | ...; Default: null)
|esp-enc-key (string/string; Default: )||Incoming-encryption-key/outgoing-encryption-key|
|esp-spi (0x100..FFFFFFFF/0x100..FFFFFFFF; Default: 0x100)||Incoming-SA-SPI/outgoing-SA-SPI|
|lifetime (time; Default: 0s)||Lifetime of this SA|
|name (string; Default: )||Name of the item for reference from policies|
/ip ipsec installed-sa
This facility provides information about installed security associations including the keys.
Sometimes after incorrect/incomplete negotiations took place, it is required to flush manually the installed SA table so that SA could be renegotiated. This option is provided by the
/ip ipsec installed-sa flush command.
This command accepts only one property:
|sa-type (ah | all | esp; Default: all)||Specifies SA types to flush.
/ip ipsec remote-peers
This submenu provides you with various statistics about remote peers that currently have established phase 1 connections with this router. Note that if peer doesn't show up here, it doesn't mean that no IPsec traffic is being exchanged with it.
Read only properties:
|local-address (ip/ipv6 address)||Local ISAKMP SA address on the router used by the peer|
|remote-address (ip/ipv6 address)||Remote peer's ip/ipv6 address|
|side (initiator | responder)||Shows which side initiated the Phase1 negotiation.|
|state (string)||State of phase 1 negotiation with the peer. For example when phase1 and phase 2 are negotiated it will show state "established".|
|established (time)||How long peers are in established state.|
Closing all IPsec connections
Menu has a command to quickly close all established ipsec connections. This command will clear all installed SAs (Phase2) and remove all entries from remote-peers menu (Phase1).
/ip ipsec remote-peers kill-connections
/ip ipsec statistics
This menu shows various ipsec statistics
|in-errors (integer)||All inbound errors that are not matched by other counters.|
|in-buffer-errors (integer)||No free buffer.|
|in-header-errors (integer)||Header error|
|in-no-states (integer)||No state is found i.e. Either inbound SPI, address, or IPsec protocol at SA is wrong|
|in-state-protocol-errors (integer)||Transformation protocol specific error, for example SA key is wrong or hardware accelerator is unable to handle amount of packets.|
|in-state-mode-errors (integer)||Transformation mode specific error|
|in-state-sequence-errors (integer)||Sequence number is out of window|
|in-state-expired (integer)||State is expired|
|in-state-mismatches (integer)||State has mismatched option, for example UDP encapsulation type is mismatched.|
|in-state-invalid (integer)||State is invalid|
|in-template-mismatches (integer)||No matching template for states, e.g. Inbound SAs are correct but SP rule is wrong|
|in-no-policies (integer)||No policy is found for states, e.g. Inbound SAs are correct but no SP is found|
|in-policy-blocked (integer)||Policy discards|
|in-policy-errors (integer)||Policy errors|
|out-errors (integer)||All outbound errors that are not matched by other counters|
|out-bundle-errors (integer)||Bundle generation error|
|out-bundle-check-errors (integer)||Bundle check error|
|out-no-states (integer)||No state is found|
|out-state-protocol-errors (integer)||Transformation protocol specific error|
|out-state-mode-errors (integer)||Transformation mode specific error|
|out-state-sequence-errors (integer)||Sequence errors, for example Sequence number overflow|
|out-state-expired (integer)||State is expired|
|out-policy-blocked (integer)||Policy discards|
|out-policy-dead (integer)||Policy is dead|
|out-policy-errors (integer)||Policy error|
Site to Site IpSec Tunnel
Consider setup as illustrated below
Two remote office routers are connected to internet and office workstations behind routers are NATed. Each office has its own local subnet, 10.1.202.0/24 for Office1 and 10.1.101.0/24 for Office2. Both remote offices needs secure tunnel to local networks behind routers.
On both routers ether1 is used as wan port and ether2 is used to connect workstations. Also NAT rules are set tu masquerade local networks.
/ip address add address=192.168.90.1/24 interface=ether1 add address=10.1.202.1/24 interface=ether2 /ip route add gateway=192.168.90.254 /ip firewall nat add chain=srcnat out-interface=ether1 action=masquerade
/ip address add address=192.168.80.1/24 interface=ether1 add address=10.1.101.1/24 interface=ether2 /ip route add gateway=192.168.80.254 /ip firewall nat add chain=srcnat out-interface=ether1 action=masquerade
IpSec Peer's config
Next step is to add peer's configuration. We need to specify peers address and port and pre-shared-key. Other parameters are left to default values.
/ip ipsec peer add address=192.168.80.1/32 port=500 auth-method=pre-shared-key secret="test"
/ip ipsec peer add address=192.168.90.1/32 port=500 auth-method=pre-shared-key secret="test"
Policy and proposal
It is important that proposed authentication and encryption algorithms match on both routers. In this example we can use predefined "default" proposal
[admin@MikroTik] /ip ipsec proposal> print Flags: X - disabled 0 name="default" auth-algorithms=sha1 enc-algorithms=3des lifetime=30m pfs-group=modp1024
As we already have proposal as a next step we need correct IpSec policy. We want to encrypt traffic coming form 10.1.202.0/24 to 10.1.101.0/24 and vice versa.
/ip ipsec policy add src-address=10.1.202.0/24 src-port=any dst-address=10.1.101.0/24 dst-port=any \ sa-src-address=192.168.90.1 sa-dst-address=192.168.80.1 \ tunnel=yes action=encrypt proposal=default
/ip ipsec policy add src-address=10.1.101.0/24 src-port=any dst-address=10.1.202.0/24 dst-port=any \ sa-src-address=192.168.80.1 sa-dst-address=192.168.90.1 \ tunnel=yes action=encrypt proposal=default
Note that we configured tunnel mode instead of transport, as this is site to site encryption.
At this point if you will try to establish IpSec tunnel it will not work, packets will be rejected. This is because both routers have NAT rules that is changing source address after packet is encrypted. Remote router reiceves encrypted packet but is unable to decrypt it because source address do not match address specified in policy configuration. For more information see packet flow ipsec example.
To fix this we need to set up NAT bypass rule.
/ip firewall nat add chain=srcnat action=accept place-before=0 \ src-address=10.1.202.0/24 dst-address=10.1.101.0/24
/ip firewall nat add chain=srcnat action=accept place-before=0 \ src-address=10.1.101.0/24 dst-address=10.1.202.0/24
It is very important that bypass rule is placed at the top of all other NAT rules.