RFC 8907: The Terminal Access Controller Access-Control System Plus (TACACS+) Protocol
- T. Dahm,
- A. Ota,
- D.C. Medway Gash,
- D. Carrel,
- L. Grant
This RFC was updated
Abstract
This document describes the Terminal Access Controller Access-Control System Plus (TACACS+) protocol, which is widely deployed today to provide Device Administration for routers, network access servers, and other networked computing devices via one or more centralized servers.¶
Status of This Memo
This document is not an Internet Standards Track specification; it is published for informational purposes.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Not all documents approved by the IESG are candidates for any level of Internet Standard; see Section 2 of RFC 7841.¶
Information about the current status of this document, any
errata, and how to provide feedback on it may be obtained at
https://
Copyright Notice
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(https://
This document may contain material from IETF Documents or IETF Contributions published or made publicly available before November 10, 2008. The person(s) controlling the copyright in some of this material may not have granted the IETF Trust the right to allow modifications of such material outside the IETF Standards Process. Without obtaining an adequate license from the person(s) controlling the copyright in such materials, this document may not be modified outside the IETF Standards Process, and derivative works of it may not be created outside the IETF Standards Process, except to format it for publication as an RFC or to translate it into languages other than English.¶
1. Introduction
This document describes the Terminal Access Controller Access-Control System Plus (TACACS+) protocol. It was conceived initially as a general Authentication, Authorization, and Accounting (AAA) protocol. It is widely deployed today but is mainly confined for a specific subset of AAA called Device Administration, which includes authenticating access to network devices, providing central authorization of operations, and auditing of those operations.¶
A wide range of TACACS+ clients and servers is already deployed in the field. The TACACS+ protocol they are based on is defined in a document that was originally intended for IETF publication, but was never standardized. The document is known as "The Draft" [THE-DRAFT].¶
This Draft was a product of its time and did not address all of the key security concerns that are considered when designing modern standards. Therefore, deployment must be executed with care. These concerns are addressed in Section 10.¶
The primary intent of this informational document is to clarify the subset of "The Draft", which is common to implementations supporting Device Administration. It is intended that all implementations that conform to this document will conform to "The Draft". However, it is not intended that all implementations that conform to "The Draft" will conform to this document. The following features from "The Draft" have been removed:¶
The TACACS+ protocol allows for arbitrary length and content authentication exchanges to support alternative authentication mechanisms. It is extensible to provide for site customization and future development features, and it uses TCP to ensure reliable delivery. The protocol allows the TACACS+ client to request fine-grained access control and allows the server to respond to each component of that request.¶
The separation of authentication, authorization, and accounting is a key element of the design of TACACS+ protocol. Essentially, it makes TACACS+ a suite of three protocols. This document will address each one in separate sections. Although TACACS+ defines all three, an implementation or deployment is not required to employ all three. Separating the elements is useful for the Device Administration use case, specifically, for authorization and accounting of individual commands in a session. Note that there is no provision made at the protocol level to associate authentication requests with authorization requests.¶
2. Conventions
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
3. Technical Definitions
This section provides a few basic definitions that are applicable to this document.¶
3.1. Client
The client is any device that initiates TACACS+ protocol requests to mediate access, mainly for the Device Administration use case.¶
3.2. Server
The server receives TACACS+ protocol requests and replies according to its business model in accordance with the flows defined in this document.¶
3.3. Packet
All uses of the word packet in this document refer to TACACS+ protocol data units unless explicitly noted otherwise. The informal term "packet" has become an established part of the definition.¶
3.4. Connection
TACACS+ uses TCP for its transport. TCP Server port 49 is allocated by IANA for TACACS+ traffic.¶
3.5. Session
The concept of a session is used throughout this document. A TACACS+ session is a single authentication sequence, a single authorization exchange, or a single accounting exchange.¶
An accounting and authorization session will consist of a single pair of packets (the request and its reply). An authentication session may involve an arbitrary number of packets being exchanged. The session is an operational concept that is maintained between the TACACS+ client and server. It does not necessarily correspond to a given user or user action.¶
3.6. Treatment of Enumerated Protocol Values
This document describes various
enumerated values in the packet header
and the headers for specific packet
types. For example, in the
authentication start packet type, this
document defines the action field with
three values: TAC
If the server does not implement one of the defined options in a packet that it receives, or it encounters an option that is not listed in this document for a header field, then it should respond with an ERROR and terminate the session. This will allow the client to try a different option.¶
If an error occurs but the type of the incoming packet cannot be determined, a packet with the identical cleartext header but with a sequence number incremented by one and the length set to zero MUST be returned to indicate an error.¶
3.7. Treatment of Text Strings
The TACACS+ protocol makes extensive use of text strings. "The Draft" intended that these strings would be treated as byte arrays where each byte would represent a US-ASCII character.¶
More recently, server implementations have been extended to
interwork with external identity services, and so a more nuanced
approach is needed. Usernames MUST be encoded and
handled using the Username
Where specifically mentioned, data fields contain arrays of arbitrary bytes as required for protocol processing. These are not intended to be made visible through user interface to users.¶
All other text fields in TACACS+ MUST be treated as printable byte arrays of US-ASCII as defined by [RFC0020]. The term "printable" used here means the fields MUST exclude the "Control Characters" defined in Section 5.2 of [RFC0020].¶
4. TACACS+ Packets and Sessions
4.1. The TACACS+ Packet Header
All TACACS+ packets begin with the following 12-byte header. The header describes the remainder of the packet:¶
The following general rules apply to all TACACS+ packet types:¶
major_version¶
minor_version¶
type¶
seq_no¶
flags¶
session_id¶
length¶
4.2. The TACACS+ Packet Body
The TACACS+ body types are defined in the packet header. The next sections of this document will address the contents of the different TACACS+ bodies.¶
4.3. Single Connection Mode
Single Connection Mode is intended to improve performance where there is a lot of traffic between a client and a server by allowing the client to multiplex multiple sessions on a single TCP connection.¶
The packet header contains the
TAC
The client sets this flag to indicate that it supports multiplexing TACACS+ sessions over a single TCP connection. The client MUST NOT send a second packet on a connection until single-connect status has been established.¶
To indicate it will support Single Connection Mode, the server sets this flag in the first reply packet in response to the first request from a client. The server may set this flag even if the client does not set it, but the client may ignore the flag and close the connection after the session completes.¶
The flag is only relevant for the first two packets on a connection, to allow the client and server to establish Single Connection Mode. No provision is made for changing Single Connection Mode after the first two packets; the client and server MUST ignore the flag after the second packet on a connection.¶
If Single Connection Mode has not been established in the first two packets of a TCP connection, then both the client and the server close the connection at the end of the first session.¶
The client negotiates Single Connection Mode to improve efficiency. The server may refuse to allow Single Connection Mode for the client. For example, it may not be appropriate to allocate a long-lasting TCP connection to a specific client in some deployments. Even if the server is configured to permit Single Connection Mode for a specific client, the server may close the connection. For example, a server MUST be configured to time out a Single Connection Mode TCP connection after a specific period of inactivity to preserve its resources. The client MUST accommodate such closures on a TCP session even after Single Connection Mode has been established.¶
The TCP connection underlying the Single Connection Mode will close eventually either because of the timeout from the server or from an intermediate link. If a session is in progress when the client detects disconnect, then the client should handle it as described in "Session Completion" (Section 4.4). If a session is not in progress, then the client will need to detect this and restart the Single Connection Mode when it initiates the next session.¶
4.4. Session Completion
The REPLY packets defined for the packet types in the sections
below
The server responds with a PASS or a FAIL to indicate that the processing of the request completed and that the client can apply the result (PASS or FAIL) to control the execution of the action that prompted the request to be sent to the server.¶
The server responds with an ERROR to indicate that the processing of the request did not complete. The client cannot apply the result, and it MUST behave as if the server could not be connected to. For example, the client tries alternative methods, if they are available, such as sending the request to a backup server or using local configuration to determine whether the action that prompted the request should be executed.¶
Refer to "Aborting an Authentication Session" (Section 5.4.3) for details on handling additional status options.¶
When the session is complete, the TCP connection should be handled as follows, according to whether Single Connection Mode was negotiated:¶
It is recommended that client implementations provide robust schemes for dealing with servers that cannot be connected to. Options include providing a list of servers for redundancy and an option for a local fallback configuration if no servers can be reached. Details will be implementation specific.¶
The client should manage connections and handle the case of a server that establishes a connection but does not respond. The exact behavior is implementation specific. It is recommended that the client close the connection after a configurable timeout.¶
4.5. Data Obfuscation
The body of packets may be obfuscated. The following sections describe the obfuscation method that is supported in the protocol. In "The Draft", this process was actually referred to as Encryption, but the algorithm would not meet modern standards and so will not be termed as encryption in this document.¶
The obfuscation mechanism relies on a secret key, a shared secret value that is known to both the client and the server. The secret keys MUST remain secret.¶
Server implementations MUST allow a unique secret key to be associated with each client. It is a site-dependent decision as to whether or not the use of separate keys is appropriate.¶
The flag field MUST be configured with
TAC
The packet body can then be de-obfuscated by XORing it bytewise with a pseudo-random pad.¶
The pad is generated by concatenating a series of MD5 hashes (each 16 bytes long) and truncating it to the length of the input data. Whenever used in this document, MD5 refers to the "RSA Data Security, Inc. MD5 Message-Digest Algorithm" as specified in [RFC1321].¶
The first MD5 hash is generated by concatenating the session_id, the secret key, the version number, and the sequence number, and then running MD5 over that stream. All of those input values are available in the packet header, except for the secret key, which is a shared secret between the TACACS+ client and server.¶
The version number and session_id are extracted from the header.¶
Subsequent hashes are generated by using the same input stream but concatenating the previous hash value at the end of the input stream.¶
When a server detects that the secrets it has configured for the device do not match, it MUST return ERROR. For details of TCP connection handling on ERROR, refer to "Session Completion" (Section 4.4).¶
This option is deprecated and
MUST NOT be used in
production. In this case, the entire
packet body is in cleartext. A request
MUST be dropped if
TAC
After a packet body is de-obfuscated, the lengths of the component values in the packet are summed. If the sum is not identical to the cleartext datalength value from the header, the packet MUST be discarded and an ERROR signaled. For details of TCP connection handling on ERROR, refer to "Session Completion" (Section 4.4).¶
Commonly, such failures are seen when the keys are mismatched between the client and the TACACS+ server.¶
5. Authentication
Authentication is the action of determining who a user (or entity)
is. Authentication can take many forms. Traditional authentication
employs a name and a fixed password. However, fixed passwords are
vulnerable security, so many modern authentication mechanisms utilize
"one-time" passwords or a challenge
Authentication is not mandatory; it is a site-configured option. Some sites do not require it. Others require it only for certain services (see "Authorization" (Section 6)). Authentication may also take place when a user attempts to gain extra privileges and must identify himself or herself as someone who possesses the required information (passwords, etc.) for those privileges.¶
5.1. The Authentication START Packet Body
Packet fields are as follows:¶
action¶
priv_lvl¶
authen_type¶
authen_service¶
user, user_len¶
port, port_len¶
rem_addr, rem_addr_len¶
data, data_len¶
5.2. The Authentication REPLY Packet Body
The TACACS+ server sends only one type of authentication packet (a REPLY packet) to the client.¶
status¶
flags¶
server_msg, server_msg_len¶
data, data_len¶
5.3. The Authentication CONTINUE Packet Body
This packet is sent from the client to the server following the receipt of a REPLY packet.¶
user_msg, user_msg_len¶
data, data_len¶
flags¶
5.4. Description of Authentication Process
The action, authen_type, and authen_service fields (described above) combine to indicate what kind of authentication is to be performed. Every authentication START, REPLY, and CONTINUE packet includes a data field. The use of this field is dependent upon the kind of authentication.¶
This document defines a core set of authentication flows to be supported by TACACS+. Each authentication flow consists of a START packet. The server responds either with a request for more information (GETDATA, GETUSER, or GETPASS) or a termination PASS, FAIL, ERROR, or RESTART. The actions and meanings when the server sends a RESTART or ERROR are common and are described further below.¶
When the REPLY status equals TAC
The client should interpret TAC
If the information being requested by the server from the client is
sensitive, then the server should set the TAC
The data field is only used in the REPLY where explicitly defined below.¶
5.4.1. Version Behavior
The TACACS+ protocol is versioned to allow revisions while maintaining backwards compatibility. The version number is in every packet header. The changes between minor_version 0 and 1 apply only to the authentication process, and all deal with the way that Challenge Handshake Authentication Protocol (CHAP) and Password Authentication Protocol (PAP) authentications are handled. minor_version 1 may only be used for authentication kinds that explicitly call for it in the table below:¶
The '-' symbol represents that the option is not valid.¶
All authorization and accounting and ASCII authentication use minor_version 0.¶
PAP, CHAP, and MS-CHAP login use minor_version 1. The normal exchange is a single START packet from the client and a single REPLY from the server.¶
The removal of SENDPASS was prompted by security concerns and is no longer considered part of the TACACS+ protocol.¶
5.4.2. Common Authentication Flows
This section describes common authentication flows. If the server does
not implement an option, it MUST respond with
TAC
5.4.2.1. ASCII Login
This is a standard ASCII authentication. The START packet
MAY contain the username. If the user does not include
the username, then the server MUST obtain it from the
client with a CONTINUE TAC
5.4.2.2. PAP Login
The entire exchange MUST consist of a single START packet and a single REPLY. The START packet MUST contain a username and the data field MUST contain the PAP ASCII password. A PAP authentication only consists of a username and password [RFC1334] (Obsolete). The REPLY from the server MUST be either a PASS, FAIL, or ERROR.¶
5.4.2.3. CHAP Login
The entire exchange MUST consist of a single START packet and a single REPLY. The START packet MUST contain the username in the user field, and the data field is a concatenation of the PPP id, the challenge, and the response.¶
The length of the challenge value can be determined from the length of the data field minus the length of the id (always 1 octet) and the length of the response field (always 16 octets).¶
To perform the authentication, the server calculates the PPP hash as defined in PPP Authentication [RFC1334] and then compares that value with the response. The MD5 algorithm option is always used. The REPLY from the server MUST be a PASS, FAIL, or ERROR.¶
The selection of the challenge and its length are not an aspect of the
TACACS+ protocol. However, it is strongly recommended that the
client
5.4.2.4. MS-CHAP v1 Login
The entire exchange MUST consist of a single START packet and a single REPLY. The START packet MUST contain the username in the user field, and the data field will be a concatenation of the PPP id, the MS-CHAP challenge, and the MS-CHAP response.¶
The length of the challenge value can be determined from the length of the data field minus the length of the id (always 1 octet) and the length of the response field (always 49 octets).¶
To perform the authentication, the server will use a combination of MD4 and DES on the user's secret and the challenge, as defined in [RFC2433], and then compare the resulting value with the response. The REPLY from the server MUST be a PASS or FAIL.¶
For best practices, please refer to [RFC2433]. The TACACS+ server MUST reject authentications where the challenge deviates from 8 bytes as defined in the RFC.¶
5.4.2.5. MS-CHAP v2 Login
The entire exchange MUST consist of a single START packet and a single REPLY. The START packet MUST contain the username in the user field, and the data field will be a concatenation of the PPP id, the MS-CHAP challenge, and the MS-CHAP response.¶
The length of the challenge value can be determined from the length of the data field minus the length of the id (always 1 octet) and the length of the response field (always 49 octets).¶
To perform the authentication, the server will use the algorithm specified [RFC2759] on the user's secret and challenge, and then compare the resulting value with the response. The REPLY from the server MUST be a PASS or FAIL.¶
For best practices for MS-CHAP v2, please refer to [RFC2759]. The TACACS+ server MUST reject authentications where the challenge deviates from 16 bytes as defined in the RFC.¶
5.4.2.6. Enable Requests
This is an "ENABLE" request, used to change the current running privilege level of a user. The exchange MAY consist of multiple messages while the server collects the information it requires in order to allow changing the principal's privilege level. This exchange is very similar to an ASCII login (Section 5.4.2.1).¶
In order to readily distinguish "ENABLE" requests from other types of request,
the value of the authen_service field MUST be set to
TAC
5.4.2.7. ASCII Change Password Request
This exchange consists of multiple messages while the server collects
the information it requires in order to change the user's password. It
is very similar to an ASCII login. The status value
TAC
5.4.3. Aborting an Authentication Session
The client may prematurely terminate a session by setting the
TAC
In cases of PASS, FAIL, or ERROR, the server can insert a message into server_msg to be displayed to the user.¶
"The Draft" [THE-DRAFT]
defined a mechanism to direct authentication requests to an
alternative server. This mechanism is regarded as insecure, is
deprecated, and is not covered here. The client should treat
TAC
If the status equals TAC
If the status equals TAC
If a client does not implement the TAC
6. Authorization
In the TACACS+ protocol, authorization is the action of determining what a user is allowed to do. Generally, authentication precedes authorization, though it is not mandatory that a client use the same service for authentication that it will use for authorization. An authorization request may indicate that the user is not authenticated (we don't know who they are). In this case, it is up to the server to determine, according to its configuration, if an unauthenticated user is allowed the services in question.¶
Authorization does not merely provide yes or no answers, but it may also customize the service for the particular user. A common use of authorization is to provision a shell session when a user first logs into a device to administer it. The TACACS+ server might respond to the request by allowing the service, but placing a time restriction on the login shell. For a list of common arguments used in authorization, see "Authorization Arguments" (Section 8.2).¶
In the TACACS+ protocol, an authorization is always a single pair of messages: a REQUEST from the client followed by a REPLY from the server.¶
The authorization REQUEST message contains a fixed set of fields that indicate how the user was authenticated and a variable set of arguments that describe the services and options for which authorization is requested.¶
The REPLY contains a variable set of response arguments (argument-value pairs) that can restrict or modify the client's actions.¶
7. Accounting
Accounting is typically the third action after authentication and
authorization. But again, neither authentication nor authorization is
required. Accounting is the action of recording what a user is doing
and/or has done. Accounting in TACACS+ can serve two purposes: it may
be used as an auditing tool for security services, and it may also be
used to account for services used such as in a billing
environment. To this end, TACACS+ supports three types of accounting
records: Start records indicate that a service is about to begin, Stop
records indicate that a service has just terminated, and Update
records are intermediate notices that indicate that a service is still
being performed. TACACS+ accounting records contain all the
information used in the authorization records and also contain
accounting
7.1. The Account REQUEST Packet Body
flags¶
All other fields are defined in "Authentication
See "Accounting Arguments" (Section 8.3) for the dictionary of arguments relevant to accounting.¶
7.2. The Accounting REPLY Packet Body
The purpose of accounting is to record the action that has occurred on the client. The server MUST reply with success only when the accounting request has been recorded. If the server did not record the accounting request, then it MUST reply with ERROR.¶
status¶
server_msg, server_msg_len¶
data, data_len¶
TACACS+ accounting is intended to record various types of events on clients, for example: login sessions, command entry, and others as required by the client implementation. These events are collectively referred to in "The Draft" [THE-DRAFT] as "tasks".¶
The TAC
The START and STOP flags are mutually exclusive.¶
The WATCHDOG flag is used by the client to communicate ongoing status of a long-running task. Update records are sent at the client's discretion. The frequency of the update depends upon the intended application: a watchdog to provide progress indication will require higher frequency than a daily keep-alive. When the WATCHDOG flag is set along with the START flag, it indicates that the update record provides additional or updated arguments from the original START record. If the START flag is not set, then this indicates only that task is still running, and no new information is provided (servers MUST ignore any arguments). The STOP flag MUST NOT be set in conjunction with the WATCHDOG flag.¶
The server MUST respond
with TAC
8. Argument-Value Pairs
TACACS+ is intended to be an extensible protocol. The arguments used in Authorization and Accounting are not limited by this document. Some arguments are defined below for common use cases. Clients MUST use these arguments when supporting the corresponding use cases.¶
8.1. Value Encoding
All argument values are encoded as strings. For details of text encoding, see "Treatment of Text Strings" (Section 3.7). The following type representations SHOULD be followed.¶
Numeric¶
Boolean¶
IP-Address¶
Date Time¶
String¶
Empty Values¶
8.2. Authorization Arguments
service (String)¶
protocol (String)¶
cmd (String)¶
cmd-arg (String)¶
acl (Numeric)¶
inacl (String)¶
outacl (String)¶
addr (IP-Address)¶
addr-pool (String)¶
timeout (Numeric)¶
idletime (Numeric)¶
autocmd (String)¶
noescape (Boolean)¶
nohangup (Boolean)¶
priv-lvl (Numeric)¶
8.3. Accounting Arguments
The following arguments are defined for TACACS+ accounting only. They MUST precede any argument-value pairs that are defined in "Authorization" (Section 6).¶
task_id (String)¶
start_time (Date Time)¶
stop_time (Date Time)¶
elapsed_time (Numeric)¶
timezone (String)¶
event (String)¶
reason (String)¶
bytes (Numeric)¶
bytes_in (Numeric)¶
bytes_out (Numeric)¶
paks (Numeric)¶
paks_in (Numeric)¶
paks_out (Numeric)¶
err_msg (String)¶
Where the TACACS+ deployment is used to support the Device Administration use case, it is often required to log all commands entered into client devices. To support this mode of operation, TACACS+ client devices MUST be configured to send an accounting start packet for every command entered, irrespective of how the commands were authorized. These "Command Accounting" packets MUST include the "service" and "cmd" arguments, and if needed, the "cmd-arg" arguments detailed in Section 8.2.¶
9. Privilege Levels
The TACACS+ protocol supports flexible authorization schemes through the extensible arguments.¶
The privilege levels scheme is built into the protocol and has been extensively used as an option for Session-based shell authorization. Privilege levels are ordered values from 0 to 15 with each level being a superset of the next lower value. Configuration and implementation of the client will map actions (such as the permission to execute specific commands) to different privilege levels. The allocation of commands to privilege levels is highly dependent upon the deployment. Common allocations are as follows:¶
A privilege level can be assigned to a shell (exec) session when it
starts. The client will permit the actions associated with this level to be
executed. This privilege level is returned by the server in a session-based
shell authorization (when "service" equals "shell" and "cmd" is empty). When
a user is required to perform actions that are mapped to a higher privilege
level, an ENABLE-type reauthenticatio
The use of privilege levels to determine session-based access to commands and resources is not mandatory for clients. Although the privilege-level scheme is widely supported, its lack of flexibility in requiring a single monotonic hierarchy of permissions means that other session-based command authorization schemes have evolved. However, it is still common enough that it SHOULD be supported by servers.¶
10. Security Considerations
"The Draft" [THE-DRAFT] from 1998 did not address all of the key security concerns that are considered when designing modern standards. This section addresses known limitations and concerns that will impact overall security of the protocol and systems where this protocol is deployed to manage central authentication, authorization, or accounting for network Device Administration.¶
Multiple implementations of the protocol described in
"The Draft" [THE-DRAFT]
have been deployed. As the protocol was never standardized, current
implementations may be incompatible in non-obvious ways, giving rise
to additional security risks. This section does not claim to enumerate
all possible security vulnerabilities
10.1. General Security of the Protocol
The TACACS+ protocol does not include a security mechanism that would meet modern-day requirements. These security mechanisms would be best referred to as "obfuscation" and not "encryption", since they provide no meaningful integrity, privacy, or replay protection. An attacker with access to the data stream should be assumed to be able to read and modify all TACACS+ packets. Without mitigation, a range of risks such as the following are possible:¶
While the protocol provides some measure of transport privacy, it is vulnerable to at least the following attacks:¶
Even though, to the best knowledge of the authors, this method of encryption wasn't rigorously tested, enough information is available that it is best referred to as "obfuscation" and not "encryption".¶
For these reasons, users deploying the TACACS+ protocol in their environments MUST limit access to known clients and MUST control the security of the entire transmission path. Attackers who can guess the key or otherwise break the obfuscation will gain unrestricted and undetected access to all TACACS+ traffic. Ensuring that a centralized AAA system like TACACS+ is deployed on a secured transport is essential to managing the security risk of such an attack.¶
The following parts of this section
enumerate only the session
10.2. Security of Authentication Sessions
Authentication sessions SHOULD be used via a secure
transport (see "TACACS+ Best Practices" (Section 10.5)) as the man
This document deprecates the redirection mechanism using the
TAC
10.3. Security of Authorization Sessions
Authorization sessions SHOULD be used via a secure
transport (see "TACACS+ Best Practices" (Section 10.5)) as it's trivial to execute a successful
man
As an example, take the field "authen
10.4. Security of Accounting Sessions
Accounting sessions SHOULD be used via a secure
transport (see "TACACS+ Best Practices" (Section 10.5)). Although Accounting sessions are not
directly involved in authentication or authorizing operations on the
device, man
In addition to these direct manipulations, different client implementations pass a different fidelity of accounting data. Some vendors have been observed in the wild that pass sensitive data like passwords, encryption keys, and the like as part of the accounting log. Due to a lack of strong encryption with perfect forward secrecy, this data may be revealed in the future, leading to a security incident.¶
10.5. TACACS+ Best Practices
With respect to the observations about the security issues described above, a network administrator MUST NOT rely on the obfuscation of the TACACS+ protocol. TACACS+ MUST be used within a secure deployment; TACACS+ MUST be deployed over networks that ensure privacy and integrity of the communication and MUST be deployed over a network that is separated from other traffic. Failure to do so will impact overall network security.¶
The following recommendations impose restrictions on how the
protocol is applied. These restrictions were not imposed in "The
Draft". New implementations
10.5.2. Connections and Obfuscation
TACACS+ servers MUST allow the definition of individual clients. The servers MUST only accept network connection attempts from these defined known clients.¶
TACACS+ servers MUST reject connections
that have
TAC
If an invalid shared secret is detected when processing packets for a client, TACACS+ servers MUST NOT accept any new sessions on that connection. TACACS+ servers MUST terminate the connection on completion of any sessions that were previously established with a valid shared secret on that connection.¶
TACACS+ clients MUST NOT set
TAC
When a TACACS+ client receives responses from servers where:¶
the TACACS+ client MUST close the TCP
session, and process the response in the same way that a
TAC
10.5.3. Authentication
To help TACACS+ administrators select stronger authentication
options, TACACS+ servers MUST allow the administrator
to configure the server to only accept challenge
TACACS+ server administrators SHOULD enable the
option mentioned in the previous paragraph.
TACACS+ server deployments SHOULD only enable other
options (such as TAC
TACACS+ server administrators SHOULD NOT allow the
same credentials to be applied in challenge-based
TAC
10.5.4. Authorization
The authorization and accounting features are intended to provide extensibility and flexibility. There is a base dictionary defined in this document, but it may be extended in deployments by using new argument names. The cost of the flexibility is that administrators and implementers MUST ensure that the argument and value pairs shared between the clients and servers have consistent interpretation.¶
TACACS+ clients that receive an unrecognized mandatory argument
MUST evaluate server response as if they received
TAC
10.5.5. Redirection Mechanism
"The Draft" described a redirection mechanism
TACACS+ servers MUST deprecate the redirection mechanism.¶
If the redirection mechanism is implemented, then TACACS+ servers MUST disable it by default and MUST warn TACACS+ server administrators that it must only be enabled within a secure deployment due to the risks of revealing shared secrets.¶
TACACS+ clients SHOULD deprecate this feature by treating
TAC
11. IANA Considerations
This document has no IANA actions.¶
12. References
12.1. Normative References
- [RFC0020]
-
Cerf, V., "ASCII format for network interchange", STD 80, RFC 20, DOI 10
.17487 , , <https:///RFC0020 www >..rfc -editor .org /info /rfc20 - [RFC1321]
-
Rivest, R., "The MD5 Message-Digest Algorithm", RFC 1321, DOI 10
.17487 , , <https:///RFC1321 www >..rfc -editor .org /info /rfc1321 - [RFC1334]
-
Lloyd, B. and W. Simpson, "PPP Authentication Protocols", RFC 1334, DOI 10
.17487 , , <https:///RFC1334 www >..rfc -editor .org /info /rfc1334 - [RFC2119]
-
Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10
.17487 , , <https:///RFC2119 www >..rfc -editor .org /info /rfc2119 - [RFC2433]
-
Zorn, G. and S. Cobb, "Microsoft PPP CHAP Extensions", RFC 2433, DOI 10
.17487 , , <https:///RFC2433 www >..rfc -editor .org /info /rfc2433 - [RFC2759]
-
Zorn, G., "Microsoft PPP CHAP Extensions, Version 2", RFC 2759, DOI 10
.17487 , , <https:///RFC2759 www >..rfc -editor .org /info /rfc2759 - [RFC3579]
-
Aboba, B. and P. Calhoun, "RADIUS (Remote Authentication Dial In User Service) Support For Extensible Authentication Protocol (EAP)", RFC 3579, DOI 10
.17487 , , <https:///RFC3579 www >..rfc -editor .org /info /rfc3579 - [RFC4086]
-
Eastlake 3rd, D., Schiller, J., and S. Crocker, "Randomness Requirements for Security", BCP 106, RFC 4086, DOI 10
.17487 , , <https:///RFC4086 www >..rfc -editor .org /info /rfc4086 - [RFC4120]
-
Neuman, C., Yu, T., Hartman, S., and K. Raeburn, "The Kerberos Network Authentication Service (V5)", RFC 4120, DOI 10
.17487 , , <https:///RFC4120 www >..rfc -editor .org /info /rfc4120 - [RFC5952]
-
Kawamura, S. and M. Kawashima, "A Recommendation for IPv6 Address Text Representation", RFC 5952, DOI 10
.17487 , , <https:///RFC5952 www >..rfc -editor .org /info /rfc5952 - [RFC8174]
-
Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10
.17487 , , <https:///RFC8174 www >..rfc -editor .org /info /rfc8174 - [RFC8265]
-
Saint-Andre, P. and A. Melnikov, "Preparation, Enforcement, and Comparison of Internationaliz
ed Strings Representing Usernames and Passwords" , RFC 8265, DOI 10.17487 , , <https:///RFC8265 www >..rfc -editor .org /info /rfc8265
12.2. Informative References
- [KRB4]
- Miller, S., Neuman, C., Schiller, J., and J. Saltzer, "Section E.2.1: Kerberos Authentication and Authorization System", MIT Project Athena, Cambridge, Massachusetts, .
- [THE-DRAFT]
-
Carrel, D. and L. Grant, "The TACACS+ Protocol Version 1.78", Work in Progress, Internet-Draft, draft
-grant , , <https://-tacacs -02 tools >..ietf .org /html /draft -grant -tacacs -02 - [TZDB]
-
Eggert, P. and A. Olson, "Sources for Time Zone and Daylight Saving Time Data", , <https://
www >..iana .org /time -zones
Acknowledgements
The authors would like to thank the following reviewers whose comments and contributions made considerable improvements to this document: Alan DeKok, Alexander Clouter, Chris Janicki, Tom Petch, Robert Drake, and John Heasley, among many others.¶
The authors would particularly like to thank Alan DeKok, who provided significant insights and recommendations on all aspects of the document and the protocol. Alan DeKok has dedicated considerable time and effort to help improve the document, identifying weaknesses and providing remediation.¶
The authors would also like to thank the support from the OPSAWG Chairs and advisors, especially Joe Clarke.¶