TACACS+ TLS 1.3

The Terminal Access Controller Access-Control System Plus (TACACS+) Protocol provides Device Administration for routers, network access servers and other networked computing devices via one or more centralized servers. The protocol provides authentication, authorization and accounting services (AAA) for TACACS+ clients within the Device Administration use case. While the content of the protocol is highly sensitive, TACACS+ lacks effective confidentiality, integrity, and authentication of the connection and network traffic between the server and client, requiring secure transport to safeguard a deployment. The existing TACACS+ mechanisms are extremely weak as described in Section 10 of . To address these deficiencies, this document updates the TACACS+ protocol to use TLS 1.3 authentication and encryption, and obsoletes the use of its former mechanisms (Section 10.5 of ).

The terms defined in Section 3 of are fully applicable here and will not be repeated. The following terms are also used in this document.

TACACS+ was originally intended to incorporate a mechanism for securing the body of its packets. The algorithm is categorized as Obfuscation in . The term should be interpreted to mean "plaintext", it is used to ensure that the algorithm is not mistaken for encryption.

This term refers to the connection defined in . It is a connection without TLS and therefore being plaintext or possibly using unsecure TACACS+ authentication and obfuscation. The use of well-known TCP/IP server port number 49 is specified as the default for Non-TLS connections.

A TLS connection is a TCP/IP connection with TLS authentication and encryption used by TACACS+ for transport. A TLS connection for TACACS+ is always between one TACACS+ client and one TACACS+ Server.

A TACACS+ server is an instance of the server as defined in that responds to TACACS+ traffic bound to a specific port number on a host. This document distinguishes the TACACS+ server instance from the host itself. A host may have multiple TACACS+ servers running, each listening on different port numbers.

The peer of a TACACS+ client (or server) in the context of a TACACS+ connection, is a server (or client). Together, the ends of a TACACS+ connection are referred to as peers.

TACACS+ over TLS takes the protocol defined in , removes the option for MD5 obfuscation, and specifies the use of TLS (version 1.3 or later) for transport using a new well-known default server port number. The next sections provide further details and guidance. TLS is introduced into TACACS+ to fulfill the following requirements:

Confidentiality and Integrity: The MD5 obfuscation specified in has been shown to be insecure . This prevents TACACS+ being used in a compliant deployment. Securing TACACS+ protocol with TLS is intended to provide confidentiality and integrity without requiring the provision of a secured network.
Peer authentication: The authentication capabilities of TLS replace the pre-shared keys of obfuscation for mutual authentication.

All data exchanged by TACACS+ peers MUST be encrypted, including the mutual authentication of the peers. Therefore, when a TCP connection is established for the service, a TLS handshake begins immediately. To ensure separation of TACACS+ traffic that uses TLS from that which does not (), they will be deployed on different ports. Due to the widespread use of default port number settings in numerous existing TACACS+ client configurations, a well-known system TCP/IP port number will be requested from IANA. Client implementations can ensure traffic is kept separate. The designated port number is [TBD] with the service name [TBDN]. Under exceptional circumstances, this document permits any other TCP port number to be configured when required by deployment specifics, but the implications in have to be considered by operators.

A TACACS+ client initiates a TLS connection by making a TCP connection to a configured server on the TACACS+ TLS port number ([TBD]) (). Once the TCP connection is established, the client MUST immediately begin the TLS negotiation before sending any TACACS+ protocol data. Implementations MUST support TLS 1.3 and MAY permit TLS 1.3 session resumption. If resumption is supported, the resumption ticket_lifetime SHOULD be configurable, including a zero seconds lifetime. Once the TLS connection is established, the exchange of TACACS+ data proceeds as normal, except that it is transmitted over TLS as TLS application data and without TACACS+ obfuscation () The connection persists until the server or client closes it. It might be closed due to an error or at the conclusion of the TACACS+ session. If Single Connection Mode has been negotiated, it might remain open after a successful session, until an error or an inactivity timeout occurs. Why it closed has no bearing on TLS resumption, unless closed by a TLS error, in which case the ticket might be invalidated.

Implementations MUST support the TLS 1.3 mandatory cipher suites (TLS 1.3 Section 9.1). The cipher suites offered or accepted SHOULD be configurable so that operators can adapt. This document makes no cipher suite recommendations, but recommendations can be found in the TLS Cipher Suites section of the .

Certificate based mutual authentication MUST be supported. Each peer MUST validate the certificate path of the remote peer, including revocation checking, as described in . If the verification succeeds, the authentication is successful and the connection is permitted. Policy may impose further constraints upon the peer, allowing or denying the connection based on certificate fields or any other parameters exposed by the implementation. Unless disabled by configuration, a peer MUST disconnect any peer that presents an invalid TLS Certificate. In addition to full certificate-based TLS authentication, implementations MAY support:

Raw Public Keys
Pre-Shared Keys (PSKs) , also known as external PSKs in TLS 1.3, which are not resumption PSKs.

The details of these mechanisms are considered out-of-scope for this document. Please refer to the RFCs above for implementation, deployment, and security considerations.

Implementations MUST support certificate Path verification as described in . Because a peer could be isolated from a remote peer's Certification Authority (CA), implementations MUST support certificate chains (a.k.a. bundles or chains of trust), where the entire chain of the remote's certificate is stored on the local peer. TLS Cached Information Extension SHOULD be implemented. This MAY be augmented with Raw Public Keys . Other approaches may be used for loading the intermediate certificates onto the client, but MUST include support for revocation checking. For example, details the AIA (Authority Information Access) extension to provide information about the issuer of the certificate in which the extension appears. It can be used to provide the address of the Online Certificate Status Protocol (OCSP) responder from where revocation status of the certificate (which includes the extension) can be checked.

For the client-side validation of presented server identities, implementations MUST follow validation techniques. Identifier types DNS-ID, IP-ID or SRV-ID are applicable for use with the TLS TACACS+ protocol, selected by operators depending upon the deployment design. TLS TACACS+ does not use URI-IDs for server identity verification. The wildcard character MUST NOT be included in the presented server identities. For the server-side validation of client identities, Implementations must support the ability to configure which fields of a certificate are used for client identification, to verify that the client is a valid source for the received certificate and that it is permitted access to TACACS+. Implementations MUST support either: Network location based validation methods as described in . or Client Identity validation of a shared identity in the certificate subjectAltName. This is applicable in deployments where the client securely supports an identity which is shared with the server. This approach allows a client's network location to be reconfigured without issuing a new client certificate, in this case, only the server mapping needs to be updated. Implementations SHOULD support the TLS Server Name Indication extension (SNI) (), and SHOULD support the ability to configure the server's domain name, so that it may be included in the SNI "server_name" extension of the client hello. Certificate provisioning is out of scope of this document.

describes the obfuscation mechanism, documented in . Such a method is weak. The introduction of TLS PSK, certificate peer authentication, and TLS encryption to TACACS+ replaces these former mechanisms and so obfuscation is hereby obsoleted. This section describes how the TACACS+ client and servers MUST operate with regards to the obfuscation mechanism. Peers MUST NOT use obfuscation with TLS. A TACACS+ client initiating a TACACS+ TLS connection MUST set the TAC_PLUS_UNENCRYPTED_FLAG bit, thereby asserting that obfuscation is not used for the session. All subsequent packets MUST have the TAC_PLUS_UNENCRYPTED_FLAG set. A TACACS+ server that receives a packet with the TAC_PLUS_UNENCRYPTED_FLAG not set (cleared) over a TLS connection, MUST return an error of TAC_PLUS_AUTHEN_STATUS_ERROR, TAC_PLUS_AUTHOR_STATUS_ERROR, or TAC_PLUS_ACCT_STATUS_ERROR as appropriate for the TACACS+ message type, with the TAC_PLUS_UNENCRYPTED_FLAG set, and terminate the session. This behavior corresponds to that defined in Section 4.5 of Data Obfuscation for TAC_PLUS_UNENCRYPTED_FLAG or key mismatches. A TACACS+ client that receives a packet with the TAC_PLUS_UNENCRYPTED_FLAG not set (i.e., cleared), MUST terminate the session, and SHOULD log this error.

This document improves the confidentiality, integrity, and authentication of the connection and network traffic between TACACS+ peers by adding TLS support. Simply adding TLS support to the protocol does not guarantee the protection of the server and clients. It is essential for the operators and equipment vendors to adhere to the latest best practices for ensuring the integrity of network devices and selecting secure TLS key and encryption algorithms. offers substantial guidance for implementing protocols that use TLS and their deployment. Those implementing and deploying Secure TACACS+ must adhere to the recommendations relevant to TLS 1.3 outlined in , or its subsequent versions. This document outlines additional restrictions permissible under . For example, any recommendations referring to TLS 1.2, including the mandatory support, are not relevant for Secure TACACS+ as TLS 1.3 or above is mandated.

Non-TLS connections should not be used for new TACACS+ deployments. TACACS+ servers that have TLS support MUST NOT allow Non-TLS connections, because of the threat of downgrade attacks, as described in Section 5.2. Instead, separate Non-TLS TACACS+ servers can be set up to cater for these clients. Further, TLS TACACS+ servers and Non-TLS TACACS+ servers SHOULD NOT be deployed on the same host. Non-TLS connections would be better served by deploying the required Non-TLS TACACS+ servers on separate hosts. It is NOT RECOMMENDED to deploy TACACS+ without TLS authentication and encryption, including TLS using the NULL algorithm, except for within test and debug environments. Also see .

TLS 1.3 resumption and PSK techniques make it possible to send Early Data, aka. 0-RTT data, data that is sent before the TLS handshake completes. Replay of this data is a risk. Given the sensitivity of TACACS+ data, clients MUST NOT send data until the full TLS handshake completes; that is, clients MUST NOT send 0-RTT data and servers MUST abruptly disconnect clients that do.

Recommendations in Section 4 of or any RFCs which obsolete it, MUST be followed, in order to determine which TLS versions and algorithms should be supported, deprecated, obsoleted, or abandoned. Other useful examples are the TLS specifications themselves (TLS 1.3), which prescribes mandatory support in Section 9, and TLS Recommendations .

Operators SHOULD be cognizant of the potential of server and/or client isolation from their peer's CA by network failures. Isolation from a public key certificate's CA will cause the verification of the certificate to fail and thus TLS authentication of the peer to fail. The approach mentioned in can help address this, and should be considered where implemented.

Operators SHOULD be aware that the TLS SNI extension is part of the TLS client hello, and is therefore subject to eavesdropping. Also see . If TLS Encrypted Client Hello becomes standardized and applicable to TLS 1.3, then it SHOULD be included in Secure TACACS+ implementation.

Implementors MUST ensure that the configuration scheme introduced for enabling TLS is straightforward and leaves no room for ambiguity regarding whether TLS or Non-TLS will be used between the TACACS+ client and the TACACS+ server. This document recommends the use of a separate port number that TLS enabled TACACS+ servers will listen to. Where deployments have not overridden the defaults explicitly, TACACS+ client implementations MUST use the correct values:

for Non-TLS connection TACACS+: Port 49.
for TLS connection TACACS+: (TBD).

Implementors MAY offer a single option for TACACS+ clients and servers to disable all Non-TLS TACACS+ operations. When enabled on a TACACS+ server, it will not respond to any requests from Non-TLS TACACS+ client connections. When enabled on a TACACS+ client, it will not establish any Non-TLS TACACS+ server connections.

A new port is considered appropriate and superior to a "STARTTLS" command or other negotiation method because it allows:

ease of blocking the unobfuscated or obfuscated connections by the TCP/IP port number,
passive Intrusion Detection Systems (IDSs) monitoring the unobfuscated to be unaffected by the introduction of TLS,
avoidance of Man in the Middle (MitM) attacks that can interfere with STARTTLS,
and helps prevent the accidental exposure of sensitive information due to misconfiguration.

However, co-existence of inferior authentication and obfuscated, whether an Non-TLS connection or deprecated parts that compose TLS, also presents opportunity for down-grade attacks. Causing failure of connections to the TLS-enabled service or the negotiation of shared algorithm support are two such down-grade attacks. The simplest way to address exposure from Non-TLS connection methods is to refuse Non-TLS connections at the server entirely, perhaps using separate servers for Non-TLS connections and TLS. Another approach is mutual configuration that requires TLS. Clients and servers SHOULD support configuration that requires peers, globally and individually, use TLS. Furthermore, peers SHOULD be configurable to limit offered or recognized TLS versions and algorithms to those recommended by standards bodies and implementers.

Operational and deployment considerations are spread throughout the document. While avoiding repetition, it is useful for the impatient to direct particular attention to Sections 5.2 and 5.1.5. However, it is important that the entire Section 5 is observed.

In Section 5.2, it is mentioned that for an optimal deployment of TLS TACACS+, TLS should be universally applied throughout the deployment. However, during the migration process from a Non-TLS TACACS+ deployment, operators may need to support both TLS and Non-TLS TACACS+ servers. This migration phase allows operators to gradually transition their deployments from an insecure state to a more secure one, but it is important to note that it is vulnerable to downgrade attacks. Therefore, the migration phase should be considered insecure until it is fully completed. To mitigate this hazard:

the period where any client is configured with both TLS and Non-TLS servers should be minimized.
the operator must consider the impact of mixed TLS and Non-TLS on security.

Some TACACS+ client devices in a deployment may not implement TLS. These devices will require access to Non-TLS TACACS+ servers. Operators MUST follow the recommendation of and deploy separate TACACS+ servers for these Non-TLS clients from those used for the TLS clients.

The authors request that, when this draft is accepted by the working group, the OPSAWG Chairs submit a request to IANA for an early allocation, per and , of a new well-known system TCP/IP port number for the service name "tacacss" (referenced in this document also as "TACACS+ TLS well-known port ([TBD])"), described as "TACACS+ over TLS". The service name "tacacss" follows the common practice of appending an "s" to the name given to the Non-TLS well-known port name. This allocation is justified in . This document requests IANA to add a new entity from the "Service name and Transport Protocol Port Number Registry" available at https://www.iana.org/assignments/service-name-port-numbers/ Service Name: tacacss Port Number: [TBD] Transport Protocol: TCP Description: TLS Secure Login Host Protocol (TACACSS) Assignee: IESG Contact: IETF Chair Reference: [TBDN] (This Document) RFC EDITOR: this port number should replace "[TBD]" and the service name should replace "[TBDN]" within this document. Considerations about service discovery are out of scope of this document.

The author(s) would like to thank Russ Housley, Steven M. Bellovin, Stephen Farrell, Alan DeKok, Warren Kumari, Tom Petch and Mohamed Boucadair for their support, insightful review, and/or comments. was also used as a basis for the approach to TLS.