Internet Key Exchange (IKE) for IPsec VPN

IKE Packet Processing

When a cleartext packet arrives on a Juniper Networks device that requires tunneling, and no active Phase 2 SA exists for that tunnel, Junos OS begins IKE negotiations and drops the packet. The source and destination addresses in the IP packet header are those of the local and remote IKE gateways, respectively. In the IP packet payload, there is a UDP segment encapsulating an ISAKMP (IKE) packet. The format for IKE packets is the same for Phase 1 and Phase 2. See Figure 1.

Meanwhile, the source host has sent the dropped packet again. Typically, by the time the second packet arrives, IKE negotiations are complete, and Junos OS protects the packet and all subsequent packets in the session—with IPsec before forwarding it.

The Next Payload field contains a number indicating one of the following payload types:

• 0002—SA Negotiation Payload contains a definition for a Phase 1 or Phase 2 SA.

• 0004—Proposal Payload can be a Phase 1 or Phase 2 proposal.

• 0008—Transform Payload gets encapsulated in a proposal payload that gets encapsulated in an SA payload.

• 0010—Key Exchange (KE) Payload contains information necessary for performing a key exchange, such as a DH public value.

• 0020—Identification (IDx) Payload.

  • In Phase 1, IDii indicates the initiator ID, and IDir indicates the responder ID.

  • In Phase 2, IDui indicates the user initiator, and IDur indicates the user responder.

  The IDs are IKE ID types such as FQDN, U-FQDN, IP address, and ASN.1_DN.

• 0040—Certificate (CERT) Payload.

• 0080—Certificate Request (CERT_REQ) Payload.

• 0100—Hash (HASH) Payload contains the digest output of a particular hash function.

• 0200—Signature (SIG) Payload contains a digital signature.

• 0400—Nonce (Nx) Payload contains some pseudorandom information necessary for the exchange).

• 0800—Notify Payload.

• 1000—ISAKMP Delete Payload.

• 2000—Vendor ID (VID) Payload can be included anywhere in Phase 1 negotiations. Junos OS uses it to mark support for NAT-T.

Each ISAKMP payload begins with the same generic header, as shown in Figure 2.

There can be multiple ISAKMP payloads chained together, with each subsequent payload type indicated by the value in the Next Header field. A value of 0000 indicates the last ISAKMP payload. See Figure 3 for an example.

IPsec Packet Processing

After IKE negotiations complete and the two IKE gateways have established Phase 1 and Phase 2 security associations (SAs), all subsequent packets are forwarded using the tunnel. If the Phase 2 SA specifies the Encapsulating Security Protocol (ESP) in tunnel mode, the packet looks like the one shown in Figure 4. The device adds two additional headers to the original packet that the initiating host sends.

As shown in Figure 4, the packet that the initiating host constructs includes the payload, the TCP header, and the inner IP header (IP1).

The router IP header (IP2), which Junos OS adds, contains the IP address of the remote gateway as the destination IP address and the IP address of the local router as the source IP address. Junos OS also adds an ESP header between the outer and inner IP headers. The ESP header contains information that allows the remote peer to properly process the packet when it receives it. This is shown in Figure 5.

The Next Header field indicates the type of data in the payload field. In tunnel mode, this value is 4, indicating an IP packet is contained within the payload. See Figure 6.

Configuring IKEv2 in Junos OS

A VPN peer is configured as either IKEv1 or IKEv2. When a peer is configured as IKEv2, it cannot fall back to IKEv1 if its remote peer initiates IKEv1 negotiation. By default, Juniper Networks security devices are IKEv1 peers.

Use the version v2-only configuration statement at the [edit security ike gateway gw-name] hierarchy level to configure IKEv2.

The IKE version is displayed in the output of the show security ike security-associations and show security ipsec security-associations CLI operational commands.

Juniper Networks devices support up to four proposals for Phase 2 negotiations, allowing you to define how restrictive a range of tunnel parameters you will accept. Junos OS provides predefined standard, compatible, and basic Phase 2 proposal sets. You can also define custom Phase 2 proposals.

Understanding IKEv2 Configuration Payload

Configuration payload is an Internet Key Exchange version 2 (IKEv2) option offered to propagate provisioning information from a responder to an initiator. IKEv2 configuration payload is supported with route-based VPNs only.

RFC 5996, Internet Key Exchange Protocol Version 2 (IKEv2), defines 15 different configuration attributes that can be returned to the initiator by the responder. Table 1 describes the IKEv2 configuration attributes supported on SRX Series Firewalls.

Table 1: IKEv2 Configuration Attributes

Attribute Type	Value	Description	Length
INTERNAL_IP4_ADDRESS	1	Specifies an address on the internal network. Multiple internal addresses can be requested. The responder can send up to the number of addresses requested.	0 or 4 octets
INTERNAL_IP4_NETMASK	2	Specifies the internal network's netmask value. Only one netmask value is allowed in the request and response messages (for example, 255.255.255.0), and it must be used only with an INTERNAL_IP4_ADDRESS attribute.	0 or 4 octets
INTERNAL_IP4_DNS	3	Specifies an address of a DNS server within the network. Multiple DNS servers can be requested. The responder can respond with zero or more DNS server attributes.	0 or 4 octets
INTERNAL_IP4_NBNS	4	Specifies an address of a NetBIOS name server (NBNS), for example, a WINS server, within the network. Multiple NBNS servers can be requested. The responder can respond with zero or more NBNS server attributes.	0 or 4 octets
INTERNAL_IP6_ADDRESS	8	Specifies an address on the internal network. Multiple internal addresses can be requested. The responder can send up to the number of addresses requested.	0 or 17 octets
INTERNAL_IP6_DNS	10	Specifies an address of a DNS server within the network. Multiple DNS servers can be requested. The responder can respond with zero or more DNS server attributes.	0 or 16 octets

For the IKE responder to provide the initiator with provisioning information, it must acquire the information from a specified source such as a RADIUS server. Provisioning information can also be returned from a DHCP server through a RADIUS server. On the RADIUS server, the user information should not include an authentication password. The RADIUS server profile is bound to the IKE gateway using the aaa access-profile profile-name configuration at the [edit security ike gateway gateway-name] hierarchy level.

Starting in Junos OS Release 20.3R1, on SRX5000 line with SPC3 and vSRX Virtual Firewall running iked process, we’ve improved IKEv2 configuration payload to:

• Support for IPv4 and IPv6 local address pool. You can also assign a fixed IP address to a peer.

  During IKE establishment, the initiator requests for an IPv4 address, IPv6 address, DNS address, or WINS address from the responder. After the responder has authenticated the initiator successfully, it assigns an IP address either from a local address pool or through RADIUS server. Depending on the configuration, this IP address is either assigned dynamically each time when a peer connects or assigned as a fixed IP address. If the RADIUS server responds with a framed pool, Junos OS assigns an IP address or information based on configuration from it's corresponding local pool. If you configure both local address pool and RADIUS server, the IP address allocated from RADIUS server takes precedence over the local pool. If you configure local IP address pool and the RADIUS server did not return any IP address, then local pool assigns the IP address to the request.

• Additional option, none introduced for authentication-order. See authentication-order (Access Profile).

• RADIUS accounting start and stop messages inform the state of the tunnel or peer to the RADIUS server. These messages can be used for tracking purposes or notifications to subsystems such as a DHCP server.

  Ensure that the RADIUS server support accounting start or stop messages. Also ensure that both the SRX Series Firewalls and the RADIUS server have appropriate settings to track these messages.

• Introduction of IPv6 support allows dual stack tunnels using configuration payload. During login process, IKE requests for both IPv4 and IPv6 addresses. AAA allow login only if all requested addresses have been allocated successfully. IKE terminates the negotiation if the requested IP is not allocated.

In a route-based VPN, secure tunnel (st0) interfaces operate in either point-to-multipoint or point-to-point mode. Address assignment through the IKEv2 configuration payload is now supported for point-to-multipoint or point-to-point mode. For point-to-multipoint interfaces, the interfaces must be numbered and the addresses in the configuration payload INTERNAL_IP4_ADDRESS attribute type must be within the subnetwork range of the associated point-to-multipoint interface.

Starting in Junos OS Release 20.1R1, you can configure a common password for IKEv2 configuration payload requests for an IKE gateway configuration. The common password in the range of 1 to 128 characters allows the administrator to define a common password. This password is used between the SRX Series Firewall and the RADIUS server when the SRX Series Firewall requesting an IP address on behalf of a remote IPsec peer using IKEv2 configuration payload. RADIUS server matches the credentials before it provides any IP information to the SRX Series Firewall for the configuration payload request. You can configure the common password using config-payload-password configured-password configuration statement at [edit security ike gateway gateway-name aaa access-profile access-profile-name] hierarchy level.

Both the SRX Series Firewall and the RADIUS server must have the same password configured and the radius server should be configured to use Password Authentication Protocol (PAP) as the authentication protocol. Without this, tunnel establishment will not be successful.

Figure 7 shows a typical workflow for a IKEv2 Configuration Payload.

The IKEv2 configuration payload feature is supported for both point-to-multipoint secure tunnel (st0) interfaces and point-to-point interfaces. Point-to-multipoint interfaces must be numbered, and the addresses provided in the configuration payload must be within the subnetwork range of the associated point-to-multipoint interface.

Starting in Junos OS Release 20.1R1, we support IKEv2 configuration payload feature with point-to-point interfaces on SRX5000 line and vSRX Virtual Firewall running iked.

Understanding Pico Cell Provisioning

IKEv2 configuration payload can be used to propagate provisioning information from an IKE responder, such as an SRX Series Firewall, to multiple initiators, such as LTE pico cell base stations in a cellular network. The pico cells ship from the factory with a standard configuration that allows them to connect to the SRX Series Firewall, but the pico cell provisioning information is stored on one or more provisioning servers within a protected network. The pico cells receive full provisioning information after establishing secure connections with the provisioning servers.

The workflow required to bootstrap and provision a pico cell and introduce it to service includes four distinct stages:

1. Initial addresses acquisition—The pico cell ships from the factory with the following information:

   • Configuration for the secure gateway tunnel to the SRX Series Firewall

   • Digital certificate issued by the manufacturer

   • Fully qualified domain name (FQDN) of the provisioning servers that lie within the protected network

   The pico cell boots up and acquires an address to be used for IKE negotiation from a DHCP server. A tunnel is then built to the secure gateway on the SRX Series Firewall using this address. An address for Operation, Administration, and Management (OAM) traffic is also assigned by the DHCP server for use on the protected network.

2. Pico cell provisioning—Using its assigned OAM traffic address, the pico cell requests its provisioning information—typically operator certificate, license, software, and configuration information—from servers within the protected network.

3. Reboot—The pico cell reboots and uses the acquired provisioning information to make it specific to the service provider’s network and operation model.

4. Service provision—When the pico cell enters service, it uses a single certificate that contains distinguished name (DN) and subject alternative name values with a FQDN to build two tunnels to the secure gateway on the SRX Series Firewall: one for OAM traffic and the other for Third-Generation Partnership Project (3GPP) data traffic.

Overview

With IKEv2, rekeying and reauthentication are separate processes. Rekeying establishes new keys for the IKE security association (SA) and resets message ID counters, but it does not reauthenticate the peers. Reauthentication verifies that VPN peers retain their access to authentication credentials. Reauthentication establishes new keys for the IKE SA and child SAs; rekeys of any pending IKE SA or child SA are no longer needed. After the new IKE and child SAs are created, the old IKE and child SAs are deleted.

IKEv2 reauthentication is disabled by default. You enable reauthentication by configuring a reauthentication frequency value between 1 and 100. The reauthentication frequency is the number of IKE rekeys that occurs before reauthentication occurs. For example, if the configured reauthentication frequency is 1, reauthentication occurs every time there is an IKE rekey. If the configured reauthentication frequency is 2, reauthentication occurs at every other IKE rekey. If the configured reauthentication frequency is 3, reauthentication occurs at every third IKE rekey, and so on.

You configure the reauthentication frequency with the reauth-frequency statement at the [edit security ike policy policy-name] hierarchy level. Reauthentication is disabled by setting the reauthentication frequency to 0 (the default). Reauthentication frequency is not negotiated by peers, and each peer can have its own reauthentication frequency value.

Supported Features

IKEv2 reauthentication is supported with the following features:

• IKEv2 initiators or responders

• Dead peer detection (DPD)

• Virtual routers and secure tunnel (st0) interfaces in virtual routers

• Network Address Translation traversal (NAT-T)

• Chassis clusters in active-active and active-passive mode for SRX5400, SRX5600, and SRX5800 devices

• In-service software upgrade (ISSU) on SRX5400, SRX5600, and SRX5800 devices

• Upgrade or insertion of a new Services Processing Unit (SPU) using the in-service hardware upgrade (ISHU) procedure

Limitations

Note the following caveats when using IKEv2 reauthentication:

• With NAT-T, a new IKE SA can be created with different ports from the previous IKE SA. In this scenario, the old IKE SA might not be deleted.

• In a NAT-T scenario, the initiator behind the NAT device can become the responder after reauthentication. If the NAT session expires, the NAT device might discard new IKE packets that might arrive on a different port. NAT-T keepalive or DPD must be enabled to keep the NAT session alive. For AutoVPN, we recommend that the reauthentication frequency configured on the spokes be smaller than the reauthentication frequency configured on the hub.

• Based on the reauthentication frequency, a new IKE SA can be initiated by either the initiator or the responder of the original IKE SA. Because Extensible Authentication Protocol (EAP) authentication and configuration payload require the IKE SA to be initiated by the same party as the original IKE SA, reauthentication is not supported with EAP authentication or configuration payload.

Multilevel Hierarchy for Certificate Authentication

Certificate-based authentication is an authentication method supported on SRX Series Firewalls during IKE negotiation. In large networks, multiple certificate authorities (CAs) can issue end entity (EE) certificates to their respective end devices. It is common to have separate CAs for individual locations, departments, or organizations.

When a single-level hierarchy for certificate-based authentication is employed, all EE certificates in the network must be signed by the same CA. All firewall devices must have the same CA certificate enrolled for peer certificate validation. The certificate payload sent during IKE negotiation only contains EE certificates.

Alternatively, the certificate payload sent during IKE negotiation can contain a chain of EE and CA certificates. A certificate chain is the list of certificates required to validate a peer’s EE certificate. The certificate chain includes the EE certificate and any CA certificates that are not present in the local peer.

The network administrator needs to ensure that all peers participating in an IKE negotiation have at least one common trusted CA in their respective certificate chains. The common trusted CA does not have to be the root CA. The number of certificates in the chain, including certificates for EEs and the topmost CA in the chain, cannot exceed 10.

Starting with Junos OS Release 18.1R1, validation of a configured IKE peer can be done with a specified CA server or group of CA servers. With certificate chains, the root CA must match the trusted CA group or CA server configured in the IKE policy

In the example CA hierarchy shown in Figure 8, Root-CA is the common trusted CA for all devices in the network. Root-CA issues CA certificates to the engineering and sales CAs, which are identified as Eng-CA and Sales-CA, respectively. Eng-CA issues CA certificates to the development and quality assurance CAs, which are identified as Dev-CA and Qa-CA, respectively. Host-A receives its EE certificate from Dev-CA while Host-B receives its EE certificate from Sales-CA.

Each end device needs to be loaded with the CA certificates in its hierarchy. Host-A must have Root-CA, Eng-CA, and Dev-CA certificates; Sales-CA and Qa-CA certificates are not necessary. Host-B must have Root-CA and Sales-CA certificates. Certificates can be loaded manually in a device or enrolled using the Simple Certificate Enrollment Process (SCEP).

Each end device must be configured with a CA profile for each CA in the certificate chain. The following output shows the CA profiles configured on Host-A:

The following output shows the CA profiles configured on Host-B:

DDoS Vulnerabilities on IKE Implementations

When the remote peer (initiator) sends an SA_INIT message, the local peer (responder) replies and allocates memory for the message structure. We call the session a half-open IKE session until authentication happens with the help of the IKE_AUTH message. After the peers establish an IKE security association (SA), the session becomes a full-open IKE session.

To understand the IKEv2 DDoS vulnerabilities, let's look at some of the ways an attacker can create an easy attack vector against the IKE SAs:

• Send a large number of SA_INIT messages (without IKE_AUTH messages) for which the attacker can create half-open IKE security association structures. The attack causes the device to utilize the resources and run out of memory.

• Send a large number of junk IKE_AUTH packets with the correct SPI_i and SPI_r on the initiator and the responder, respectively. The device runs out of memory while trying to decrypt the packets.

• Send SA_INIT packets continuously. The device runs out of memory while trying to generate keys for the encrypted packets.

• Send a large number of rekey requests per second during full open IKE sessions.

• Send a large number of messages with distinct message identifiers (IDs). The device queues all incoming IKE messages and runs out of memory.

How to safeguard the IKE implementations

We provide a robust infrastructure to mitigate and to monitor DDoS attacks for both IKEv1 and IKEv2 protocols. You can protect against the attacks on IKE implementations when your firewall runs the iked process (in the junos-ike package) for the IPsec VPN service.

Note:

For details about DDoS protection for IKEv2, see RFC 8019, Protecting Internet Key Exchange Protocol Version 2 (IKEv2) Implementations from Distributed Denial-of-Service Attacks. We provide similar protection for IKEv1 when your firewall runs the IPsec VPN service using the iked process. We do not support the client puzzle mechanism that the RFC presents.

Protection Against DDoS Attacks

You can enable multiple defense mechanisms against the DDoS attacks during the IKE security association creation process. These mechanisms include configuration of rate limiting and a retention period for the half-open IKE SAs, and further managing the incoming exchange rates for the rekey requests. We provide the following measures to ensure protection against DDoS attacks on IKE SAs:

• Protection measures for half-open IKE SAs:

  • The responder does not allow configuration of half-open IKE SAs for a certain duration. You can set this limit so that the responder does not configure the SAs until it reaches the timeout duration. For more details, see session (Security IKE) for the option timeout.

  • You can set a limit on the maximum allowed half-open IKE SAs on the responder. When the total number of half-open IKE SAs reaches the maximum count, the responder rejects new connections for both IKEv1 and IKEv2 SAs. For more details, see the max-count option insession (Security IKE).

  • The responder enforces threshold values on the session count for half-open IKE SAs. When the total number of half-open IKE SAs reaches the threshold values:

    • In case of IKEv2 SAs, the responder invokes cookie mechanism for any new connections.

    • In case of IKEv1 SAs, the responder rejects new connections.

    For more details, see the thresholds, send-cookie and reduce-timeout options in session (Security IKE).

  • The responder can discard duplicate sessions. For more details, see the discard-duplicate option in session (Security IKE).

  • You can set backoff timeouts for authentication failure and initiation failure phases. For more details, see session (Security IKE) for the options backoff-timeouts, init-phase-failure and auth-phase-failure.

• For full open IKE SAs:

  • You can configure maximum incoming rekey request rates to throttle the requests in a scaled scenario. For more details, see the incoming-exchange-max-rates option in session (Security IKE).

• The responder can block an incoming IKE session from peers based on the peer IKE ID. For more details, see blocklists (Security IKE).

• For dynamic gateways, you can set a limit for the number of connections at the IKE gateway configuration level using the option connections-limit. For more details, see gateway (Security IKE).

For further details on how to configure these options, see Configure Protection Against IKE DDoS Attacks.

We do not support:

• DDoS protection with IPsec VPN service based on the kmd process.

• Protection against Hash and URL certificate encoding attacks as we do not support these encoding types.

Ways to Monitor DDoS Attacks

We provide the following mechanisms to monitor the DDoS attacks:

• Use the show security ike security-associations command to list down all the matured and non-matured IKE security associations. For more details, see show security ike security-associations.

• Use the show security ike stats command to display global IKE statistics of the IPsec VPN tunnel, such as in-progress, established, expired statistics. For more details, see show security ike stats.

• Use the show security ike active-peer command to display details of the successful IKE negotiations with the remote peers. For more details, see show security ike active-peer.

• Use the show security ike peers in-progress command to display the details of in progress IKE SAs, including the half open IKE SAs. For more details, see show security ike peers. You can also see the details of the blocked, failed, and backoff peers using this command.

• Use the clear security ike peers command to clear the IKE peers that are backed off, blocked, failed or in progress. For more details, see clear security ike peers.

• To delete an existing IKE security association with peers that needs to be blocked from further communications, use the clear security ike security-associations command. For more details, see clear security ike security-associations.

• The iked system log (syslog) messages IKE_GATEWAY_PEER_BLOCKED, IKE_GATEWAY_PEER_BACKOFF and IKE_GATEWAY_PEER_FAILED provide details about the blocked, backed-off, and failed IKE negotiations with the remote peers, respectively.

• For additional security, Junos OS provides services to users for protection against packet-based system attacks such as filtering, session count, and rate limiting. For more details, see the show firewall command and the ids-option statement at the [edit security screen ids-option screen-name] hierarchy level.

Message Fragmentation

IKEv2 message fragmentation, as described in RFC 7383, Internet Key Exchange Protocol Version 2 (IKEv2) Message Fragmentation, allows IKEv2 to operate in environments where IP fragments might be blocked and peers would not be able to establish an IPsec security association (SA). IKEv2 fragmentation splits a large IKEv2 message into a set of smaller ones so that there is no fragmentation at the IP level. Fragmentation takes place before the original message is encrypted and authenticated, so that each fragment is separately encrypted and authenticated. On the receiver, the fragments are collected, verified, decrypted, and merged into the original message.

For IKEv2 fragmentation to occur, both VPN peers must indicate fragmentation support by including the IKEV2_FRAGMENTATION_SUPPORTED notification payload in the IKE_SA_INIT exchange. If both peers indicate fragmentation support, it is up to the initiator of the message exchange to determine whether or not IKEv2 fragmentation is used.

On SRX Series Firewalls, a maximum of 32 fragments are allowed per IKEv2 message. If the number of IKEv2 message fragments to be sent or received exceeds 32, the fragments are dropped and the tunnel is not established. Retransmission of individual message fragments is not supported

Configuration

On SRX Series Firewalls, IKEv2 fragmentation is enabled by default for IPv4 and IPv6 messages. To disable IKEv2 fragmentation, use the disable statement at the [edit security ike gateway gateway-name fragmentation] hierarchy level. You can also use the size statement to configure the size of the packet at which messages are fragmented; the packet size ranges from 500 to 1300 bytes. If size is not configured, the default packet size is 576 bytes for IPv4 traffic and 1280 bytes for IPv6 traffic. An IKEv2 packet that is larger than the configured packet size is fragmented.

After IKEv2 fragmentation is disabled or enabled or the packet fragment size is changed, the VPN tunnels that are hosted on the IKE gateway are brought down and IKE and IPsec SAs are renegotiated.

Caveats

The following features are not supported with IKEv2 fragmentation:

• Path MTU Discovery.

• SNMP.