| Internet-Draft | SCTP CRYPTO Chunk | September 2023 |
| Westerlund, et al. | Expires 11 March 2024 | [Page] |
This document describes a method for adding cryptographic protection to the Stream Control Transmission Protocol (SCTP). The SCTP CRYPTO chunk defined in this document is intended to enable communications privacy for applications that use SCTP as their transport protocol and allows applications to communicate in a way that is designed to prevent eavesdropping and detect tampering or message forgery.¶
The CRYPTO chunk defined here in is one half of a complete solution. Where a companion specification is required to define how the content of the CRYPTO chunk is protected, authenticated, and protected against replay, as well as how key management is accomplished.¶
Applications using SCTP CRYPTO chunk can use all transport features provided by SCTP and its extensions but with some limitations.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-westerlund-tsvwg-sctp-crypto-chunk/.¶
Discussion of this document takes place on the Transport Area Working Group (tsvwg) Working Group mailing list (mailto:[email protected]), which is archived at https://mailarchive.ietf.org/arch/browse/tsvwg/. Subscribe at https://www.ietf.org/mailman/listinfo/tsvwg/.¶
Source for this draft and an issue tracker can be found at https://github.com/gloinul/draft-westerlund-tsvwg-sctp-crypto-chunk.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 11 March 2024.¶
Copyright (c) 2023 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://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
This document defines a CRYPTO chunk for the Stream Control Transmission Protocol (SCTP), as defined in [RFC9260].¶
This specification defines the actual CRYPTO chunk. How to enable it usage, how it interacts with the SCTP association establishment to enable endpoint authentication, key-establishment, and other features require a separate protection engine specification.¶
This specification is intended to be capable of enabling mutual authentication of endpoints, data confidentiality, data origin authentication, data integrity protection, and data replay protection for SCTP packets after the SCTP association has been established. The exact properties will depend on the companion specification defining the protection engine used with the CRYPTO chunk. The protection engine specification might be based on an existing security protocol.¶
Applications using SCTP CRYPTO chunk can use most transport features provided by SCTP and its extensions. However, there can be some limitations or additional requirements for them to function such as those noted for SCTP restart and use of Dynamic Address Reconfiguration, see Section 2.8 and Section 2.9. Due to its level of integration as discussed in next section it will provide its security functions on all content of the SCTP packet, and will thus not impact the potential to utilize any SCTP functionalities or extensions that are possible to use between two SCTP peers with full security and SCTP association state.¶
The CRYPTO chunk is defined as a method for secure and confidential transfer for SCTP packets. This is implemented inside the SCTP protocol, in a sublayer between the SCTP common header handling and the SCTP chunk handling. Once an SCTP packet has been received and the SCTP common header has been used to identify the SCTP association, the CRYPTO chunk is sent to the chosen protection engine that will return the SCTP payload containing the unprotected SCTP chunks, those chunks will then be handled according to their SCTP protocol specifications. Figure 1 illustrates the CRYPTO chunk layering in regard to SCTP and the Upper Layer Protocol (ULP).¶
Use of the CRYPTO chunk is defined per SCTP association and a SCTP association uses a single protection engine. Different associations within the same SCTP endpoint may use or not use the CRYPTO chunk, and different associations may use different protection engines.¶
On the outgoing direction, once the SCTP stack has created the unprotected SCTP packet payload containing control and/or DATA chunks, that payload will be sent to the protection engine to be protected. The format of the protected payload depends on the protection engine but the unprotected payload will typically be encrypted and integrity tagged before being encapsulated in a CRYPTO chunk.¶
The SCTP protection engine performs protection operations on the whole unprotected SCTP packet payload, i.e., all chunks after the SCTP common header. Information protection is kept during the lifetime of the association and no information is sent unprotected except than the initial SCTP handshake, the SCTP common header, the SCTP CRYPTO chunk header and the SHUTDOWN-COMPLETE chunk.¶
SCTP CRYPTO chunk capability is agreed by the peers at the initialization of the SCTP association, during that phase the peers exchange information about the protection engines available. Once the SCTP association is established and the peers have agreed on what protection to use, the SCTP endpoints may start sending Protection Engine's payloads in SCTP DATA chunks containing the initialization information related to the protection engine including key agreement and endpoint authentication. This is depending on the chosen protection engine thus is not being detailed in the current specification and may be done out-of-band of the SCTP association.¶
When the endpoint authentication and key establishment has been completed, the association is considered to be secured and the ULP is informed about that. From this time on it's possible for the ULPs to exchange data securely.¶
CRYPTO chunks will never be retransmitted, retransmission is implemented by SCTP endpoint at chunk level as in the legacy. Duplicated CRYPTO chunks, whenever they will be accepted by the protection engine, will result in duplicated SCTP chunks and will be handled as duplicated chunks by SCTP endpoint in the same way a duplicated SCTP packet with those SCTP chunks would have been.¶
The protection engine, independently from the security characteristics, needs to be capable working on an unreliable transport mechanism same as UDP in regards to the payloads of the CRYPTO chunk, and have its own key management capability. SCTP is capable of providing reliable transport of key-management messages.¶
SCTP CRYPTO chunk directly exploits the protection engine by requesting protection and unprotection of a buffer, in particular the protection buffer should never exceed the possible SCTP packet size thus protection engine needs to be aware of the PMTU (see Section 2.4).¶
The key management part of the protection engine is the set of data and procedures that take care of key distribution, verification, and update. SCTP CRYPTO provides support for in-band key management, on those cases the Protection Engines uses SCTP DATA chunks identified with a dedicated Payload Protocol Identifier. The protection engine can specify if the transmission of any key-managment messages are non-reliable or reliable transmitted by SCTP.¶
During protection engine initialization, that is after the SCTP association reaches the ESTABLISHED state (see [RFC9260] Section 4), but before protection engine key-management has completed and the Protected Assocation Parameter Validation is completed, the in-band Key Management MAY use DATA chunks that SHALL use the Protection Engine PPID (see Table 9). These DATA chunks SHALL be sent unprotected by the protection engine as no keys have been established yet. As soon as the protection engine has been intialized and the validation has occured, any protection engine that uses in-band key management, i.e. sent using SCTP DATA chunks with the Protection Engine PPID, will have their message protected inside SCTP CRYPTO chunk protected with the currently established key. SCTP CRYPTO chunk state evolution is described in Section 7.3.¶
Key management MAY use other mechanism than what provided by SCTP CRYPTO chunks, in any case the mechanism for key management MUST be detailed in the specification for that protection engine.¶
The protection engines MAY exploit the Flags byte provided by the CRYPTO chunk header (see Figure 3) for its needs. Details of the use of Flags, if different from what described in the current document, MUST be specified in the Protection Engine Specification document for that specific protection engine.¶
The SCTP common header is assumed to be implicitly protected by the protection engine. This protection is based on the assumption that there will be a one-to-one mapping between SCTP association and individually established security contexts. If the protection engine does not meet that assumption further protection of the common header is likely required.¶
An example of protection engine can be DTLS as specified in [I-D.westerlund-tsvwg-sctp-crypto-dtls].¶
Protection engine and SCTP are asynchronous, meaning that the protection engine may deliver the decrypted SCTP Payload to the SCTP endpoint without respecting the reception order. It's up to SCTP endpoint to reorder the chunks in the reception buffer and to take care of the flow control according to what specified in [RFC9260]. From SCTP perspective the CRYPTO chunk processing is part of the transport network.¶
Even though the above allows the implementors to adopt a multithreading design of the protection engines, the actual implementation should consider that out-of-order handling of SCTP chunks is not desired and may cause false congestions and retransmissions.¶
The addition of the CRYPTO chunk to SCTP reduces the room for payload, due to the size of the CRYPTO chunk header and plain text expansion due to ciphering algorithm and any authentication tag. Thus, the SCTP layer creating the plain text payload needs to know about the overhead to adjust its target payload size appropriately.¶
On the other hand, the protection engine needs to be informed about the PMTU by removing from the value the sum of the common SCTP header and the CRYPTO chunk header. This implies that SCTP can propagate the computed PMTU at run time specifically. The way protection engine provides the primitive for PMTU communication shall be part of the protection engine specification.¶
From SCTP perspective, if there is a maximum size of plain text data that can be protected by the protection engine that must be communicated to SCTP. As such a limit will limit the PMTU for SCTP to the maximum plain text plus CRYPTO chunk and algorithm overhead plus the SCTP common header.¶
The SCTP mechanism for handling congestion control does depend on successful data transfer for enlarging or reducing the congestion window CWND (see [RFC9260] Section 7.2).¶
It may happen that protection engine discards packets due to internal checks or because it has detected a malicious attempt. As those packets do not represent what the peer sent, it is acceptable to ignore them, although in-situ modification on the path of a packet resulting in discarding due to integrity failure will leave a gap, but has to be accepted as part of the path behavior.¶
The protection engine shall not interfere with the SCTP congestion control mechanism, this basically means that from SCTP perspective the congestion control is exactly the same as how specified in [RFC9260].¶
The SCTP implementation will be responsible for handling ICMP messages and their validation as specified in [RFC9260] Section 10. This means that the ICMP validation needs to be done in relation to the actual sent SCTP packets with the CRYPTO chunk and not the unprotected payload. However, valid ICMP errors or information may indirectly be provided to the protection engine, such as an update to PMTU values based on packet to big ICMP messages.¶
When an Association is multihomed there are multiple paths between Endpoints. The selection of the specific path to be used at a certain time belongs to SCTP protocol that will decide according to [RFC9260]. The Protection Engine shall not influence the path selection algorithm, actually the Protection Engine will not even know what path is being used.¶
When using Dynamic Address Reconfiguration [RFC5061] in an SCTP association using CRYPTO Chunk the ASCONF chunk is protected, thus it needs to be unprotected first, furthermore it MAY come from an unknown IP Address. In order to properly address the ASCONF chunk to the relevant Association for being unprotected, Destination Address, Source and Destination ports and VTag shall be exploited. If the combination of those parameters is not unique the implementor MAY choose to send the Crypto Chunk to all Associations that fit with the parameters in order to find the right one. The association will attempt de-protection operations on the crypto chunk, and if that is successful the ASCONF chunk can be processed.¶
The recommendation [RFC5061] specifies (section 4.1.1 of [RFC5061]) that ASCONF message are required to be sent authenticated with SCTP-AUTH [RFC4895]. For SCTP associations using Crypto Chunks, when the Protection Engine provides strong Authentication such for instance in case of DTLS, results in the use of redundant mechanism for Authentication with both SCTP-AUTH and the Crypto Chunk. We recommend to amend [RFC5061] for including Crypto Chunks as Authentication mechanism for ASCONF chunks.¶
This section deals with the handling of an unexpected INIT chunk during an Association lifetime as described in [RFC9260] section 5.2 The introduction of CRYPTO CHUNK opens for two alternatives depending on if the protection engine preserves its state (crypto context) or not.¶
When the encryption engine can preserve the crypto context, meaning that encrypted data belonging to the current Association can be encrypted and decrypted, the request for SCTP Restart SHOULD use INIT chunk in CRYPTO chunk.¶
When the crypto context is not preserved, the SCTP Restart can only be accomplished by means of plain text INIT. This opens to a man-in-the-middle attack where a malicious attacker may theoretically generate an INIT chunk with proper parameters and hijack the SCTP association.¶
If the crypto context has been preserved the peer aiming for a SCTP Restart can still send CRYPTO chunks that can be processed by the remote peer. In such case the peer willing to restart the Association SHOULD send the INIT chunk in a CRYPTO chunk and encrypt it. At reception of the CRYPTO chunk containing INIT, the receiver will follow the procedure described in [RFC9260] section 5.2.2 with the exception that all the chunks will be sent in CRYPTO chunks.¶
An endpoint supporting SCTP Association Restart and implementing Crypto Chunk MUST accept receiving SCTP packets with a verification tag with value 0. The endpoint will attempt to map the packet to an association based on source IP address, destination address and port. If the combination of those parameters is not unique the implementor MAY choose to send the Crypto Chunk to all Associations that fit with the parameters in order to find the right one. Note that type of trial decrypting of the SCTP packets will increase the resource consumption per packet with the number of matching SCTP associations.¶
Note that the Association Restart will update the verification tags for both endpoints. At the end of the unexpected INIT handshaking the receiver of INIT chunk SHALL trigger the creation of a new DTLS connection to be executed as soon as possible to verify the peer identity.¶
If the crypto context isn't preserved the peer aiming for a SCTP Restart can only perform an INIT in plain text. Supporting this option opens up the SCTP association to an availability attack, where an capable attacker may be able to hijack the SCTP association. Therefore an implementation should only support and enable this option if restart is crucial.¶
To mount the attack the attacker needs to be able to process copies of packets sent from the target endpoint towards its peer for the targeted SCTP association. In addition the attacker needs to be able to send IP packets with a source address of the target's peer. If the attacker can send an SCTP INIT that appear to be from the peer, if the target is allowing this option it will generate an INIT ACK back, and assuming the attacker succesfully completes the restart handshake process the attack has managed to change the VTAG for the association and the peer will no longer respond, leading to a SCTP associatons failure.¶
As mitigation an SCTP endpoint supporting Association Restart by means of plain text INIT SHOULD support is the following. The endpoint receiving an INIT should send HEARTBEATs protected by CRYPTO CHUNK to its peer to validate that the peer is unreachable. If the endpoint receive an HEARTBEAT ACK within a reasonable time (at least a couple of RTTs) the restart INIT SHOULD be discarded as the peer obviously can respond, and thus have no need for a restart. A capable attacker can still succeed in its attack supressing the HEARTBEAT(s) through packet filtering, congestion overload or any other method preventing the HEARTBEATS or there ACKs to reach their destination. If it has been validated that the peer is unreachable, the INIT chunk will trigger the procedure described in [RFC9260] section 5.2.2¶
Note that the Association Restart will update the verification tags for both endpoints. At the end of the unexpected INIT handshaking the receiver of INIT chunk SHALL trigger the creation of a new DTLS connection to be executed as soon as possible. Also note that failure in handshaking of a new DTLS connection is considered a protocol violation and will lead to Association Abort (see Section 6.2.2).¶
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.¶
This section defines the new parameter type that will be used to negotiate the use of the CRYPTO chunk and protection engines during association setup. Table 1 illustrates the new parameter type.¶
| Parameter Type | Parameter Name |
|---|---|
| 0x80xx | Protected Association |
Note that the parameter format requires the receiver to ignore the parameter and continue processing if the parameter is not understood. This is accomplished (as described in [RFC9260], Section 3.2.1.) by the use of the upper bits of the parameter type.¶
This parameter is used to carry the list of proposed protection engines and the chosen protection engine during INIT/INIT-ACK handshake.¶
This value MUST be set to 0x80XX.¶
This value holds the length of the Protection Engines field in bytes plus 4.¶
In the INIT chunk this holds the list of protection engines in descending order of preference, i.e. the most preferred comes first, and the least preferred last in this field. In the INIT-ACK chunk this holds a single chosen protection engine. Each protection engine is specified by a 16-bit unsigned integer.¶
If the length of the Protection Engines field is not a multiple of 4 bytes, the sender MUST pad the chunk with all zero bytes to make the chunk 32-bit aligned. The Padding MUST NOT be longer than 2 bytes and it MUST be ignored by the receiver.¶
RFC-Editor Note: Please replace 0x08XX with the actual parameter type value assigned by IANA and then remove this note.¶
This section defines the new chunk type that will be used to transport protected SCTP payload. Table 2 illustrates the new chunk type.¶
| Chunk Type | Chunk Name |
|---|---|
| 0x4X | Crypto Chunk (CRYPTO) |
RFC-Editor Note: Please replace 0x4x with the actual chunk type value assigned by IANA and then remove this note.¶
It should be noted that the CRYPTO chunk format requires the receiver stop processing this SCTP packet, discard the unrecognized chunk and all further chunks, and report the unrecognized chunk in an ERROR chunk using the 'Unrecognized Chunk Type' error cause. This is accomplished (as described in [RFC9260] Section 3.2.) by the use of the upper bits of the chunk type.¶
The CRYPTO chunk is used to hold the protected payload of a plain SCTP packet.¶
This value MUST be set to 0x4X for all CRYPTO chunks.¶
This is used by the protection engine and ignored by SCTP.¶
This value holds the length of the Payload in bytes plus 4.¶
This holds the encrypted data.¶
If the length of the Payload is not a multiple of 4 bytes, the sender MUST pad the chunk with all zero bytes to make the chunk 32-bit aligned. The Padding MUST NOT be longer than 3 bytes and it MUST be ignored by the receiver.¶
This section defines the new chunk types that will be used to validate the negotiation of the protection engine selected for CRYPTO chunk. This to prevent down grade attacks on the negotiation of protection engines. Table 3 illustrates the new chunk type.¶
| Chunk Type | Chunk Name |
|---|---|
| 0x4X | Protected Association Parameter Validation (PVALID) |
It should be noted that the PVALID chunk format requires the receiver stop processing this SCTP packet, discard the unrecognized chunk and all further chunks, and report the unrecognized chunk in an ERROR chunk using the 'Unrecognized Chunk Type' error cause. This is accomplished (as described in [RFC9260] Section 3.2.) by the use of the upper bits of the chunk type.¶
The PVALID chunk is used to hold the protection engines list.¶
This value MUST be set to 0x4X.¶
MUST be set to zero on transmit and MUST be ignored on receipt.¶
This value holds the length of the Protection Engines field in bytes plus 4.¶
This holds the list of protection engines where each protection engine is specified by a 16-bit unsigned integer. The field MUST be identical to the content of the Protected Association Parameter (Section 4.1) Protection Engines field that the endpoint sent in the INIT or INIT-ACK chunk.¶
If the length of the Protection Engines field is not a multiple of 4 bytes, the sender MUST pad the chunk with all zero bytes to make the chunk 32-bit aligned. The Padding MUST NOT be longer than 2 bytes and it MUST be ignored by the receiver.¶
RFC-Editor Note: Please replace 0x4X with the actual chunk type value assigned by IANA and then remove this note.¶
This specification introduces a new set of error causes that are to be used when SCTP endpoint detects a faulty condition. The special case is when the error is detected by the protection engine that may provide additional information.¶
When an initiator SCTP endpoint sends an INIT chunk that doesn't contain the Protected Association parameter towards an SCTP endpoint that only accepts protected associations, the responder endpoint SHALL raise a Missing Mandatory Parameter error. The ERROR chunk will contain the cause code 'Missing Mandatory Parameter' (2) (see [RFC9260] Section 3.3.10.7) and the protected association parameter identifier Section 4.1 in the missing param Information field.¶
Note: Cause Length is equal to the number of missing parameters 8 + N * 2 according to [RFC9260], section 3.3.10.2. Also the Protection Association ID may be present in any of the N missing params, no order implied by the example in Figure 5.¶
A new Error Type is defined for Crypto Chunk, it's used for any error related to the Protection mechanism described in this document and has a structure that allows detailed information to be added as extra causes.¶
This specification describes some of the causes whilst the Protection Engine Specification MAY add further Causes related to the related Protection Engine.¶
When detecting an error, SCTP will send an ABORT chunk containing the relevant Error Type and Causes.¶
The SCTP Error Chunk Cause Code indicating "Error in Protection" is TBA9.¶
Is for N extra Causes equal to 4 + N * 2¶
Each Extra Cause indicate an additional piece of information as part of the error. There MAY be zero to as many as can fit in the extra cause field in the ERROR Chunk (A maximum of 32764).¶
Editor's Note: Please replace TBA9 above with what is assigned by IANA.¶
Below a number of defined Error Causes are defined, additional causes can be registered with IANA following the rules in Section 9.2.¶
If list of protection engines contained in the INIT signal doesn't contain at least an entry that fits the list of protection engines at the responder, the responder will reply with an ABORT chunk with error in protection cause code (specified in Section 6.2) and the "No Supported Protection Engine" extra cause code identifier 0x00.¶
If the protection engine specifies a handshake for example for authentication, and key management is implemented in-band, it may happen that the procedure has errors. In such case an ABORT chunk will be sent with error in protection cause code (specified in Section 6.2) and extra cause "Error During Protection Handshake" identifier 0x01.¶
A Failure may occur during protection engine Validation, i.e. when the PVALID chunks Section 5.2 are exchanged to validate the protection engine offered. In such case an ABORT chunk will be sent with error in protection cause code (specified in Section 6.2) and extra cause "Failure in Protection Engines Validation" identifier 0x02 to indicate this failure.¶
Whenever a T-valid timeout occurs, the SCTP endpoint will send an ABORT chunk with "Error in Protection" cause (specified in Section 6.2) and extra cause "Timeout During Protection Handshake or Validation" identifier 0x03 to indicate this failure. To indicate in which phase the timeout occurred an additional extra cause code is added. If the protection engine specifies that key management is implemented in-band and the T-valid timeout occurs during the handshake the Cause-Specific code to add is "Error During Protection Handshake" identifier 0x01. If the T-valid timeout occurs during the protection association parameter validation, the extra cause code to use is "Failure in Protection Engines Validation" identifier 0x02.¶
Protection engine MAY inform local SCTP endpoint about errors, in such case it's to be defined in the protection engine specification document. When an Error in the protection engine compromises the protection mechanism, the protection engine may stop processing data altogether, thus the local SCTP endpoint will not be able to send or receive any chunk for the specified Association. This will cause the Association to be closed by legacy timer-based mechanism. Since the Association protection is compromised no further data will be sent and the remote peer will also experience timeout on the Association.¶
A non-critical error in the protection engine means that the protection engine is capable of recovering without the need of the whole Association to be restarted.¶
From SCTP perspective, a non-critical error will be perceived as a temporary problem in the transport and will be handled with retransmissions and SACKS according to [RFC9260].¶
When the protection engine will experience a non-critical error, an ABORT chunk SHALL NOT be sent. This way non-critical errors are handled and how the protection engine will recover from these errors is being described in the Protection Engine Specifications.¶
An SCTP Endpoint acting as initiator willing to create a protected association shall send to the remote peer an INIT chunk containing the Protected Association parameter (see Section 4.1) where all the supported Protection Engines are listed, given in descending order of preference (see Figure 2).¶
An SCTP Endpoint acting as responder, when receiving an INIT chunk with protected association parameter, will search the list of protection engines for the most preferred commonly supported choice and will reply with INIT-ACK containing the protected association parameter with the chosen protection engine. When the responder cannot find a supported protection engine, it will reply with ABORT containing Error in Protection with the extra cause code for "No Supported Protection Engine" (Section 6.2.1).¶
Additionally, an SCTP Endpoint acting as responder willing to support only protected associations shall consider INIT chunk not containing the Protected Association parameter as an error, thus it will reply with an ABORT chunk according to what specified in Section 6.1 indicating that for this endpoint mandatory protected association parameter is missing.¶
When initiator and responder have agreed on a protected association by means of handshaking INIT/INIT-ACK with a common protection engine the SCTP association establishment continue until it has reached the ESTABLISHED state. However, before the SCTP assocation is protected by the Crypto Chunk and its protection engine some additional states needs to be passed. First the protection engine needs be initilizied in the PROTECTION INTILIZATION state. When that has been accomplished one enters the VALIDATION state where one validates the exchange of the Protected Association Parameter. If that is successful one enters the PROTECTED state. This state sequence is depicted in Section 7.3.¶
Until the procedure has reached the PROTECTED state the only usage of DATA Chunks that is accepted is DATA Chunks with the Protection Engine PPID. Any other DATA chunk being sent on a Protected association SHALL be silently discarded.¶
The Protection Engine may initialize itself by transferring its own messages as payload of the DATA chunk if necessary. The Crypto Chunk initialization SHOULD be supervised by a T-valid timer that depends on the protection engine and may also be further adjusted based on whether expected RTT values are outside of the ones commonly occurring on the general Internet, see Section 7.5. At completion of Protection Engine initialization the setup of the Protected association is complete and one enters the VALIDATION state, and from that time on only CRYPTO chunks will be exchanged. Any plain text chunk will be silently discarded.¶
If protection engine key establishment is in-band, the protection engine will start the handshake with its peer and in case of failure or T-valid timeout, the endpoint will generate an ABORT chunk. The ERROR handling follows what specified in Section 6.2.2.¶
The protection engine specification MUST specify when VALIDATION state can be entered for each endpoint. If key establishment is out-of-band, after starting T-valid timer the SCTP association will enter the VALIDATION state per protection engine specification when the necessary security context is in place.¶
When entering the VALIDATION state, the initiator MUST send to the responder a PVALID chunk (see Table 3) containing the list of Protection Engines previously sent in the protected association parameter of the INIT chunk. The transmission of the PVALID chunk MUST be done reliably. The responder receiving the PVALID chunk will compare the Protection Engines list with the one previously received in the INIT chunk, if they are exactly the same, with the same Protection engine in the same position, it will reply to the initiator with a PVALID chunk containing the chosen Protection Engine, otherwise it will reply with an ABORT chunk. ERROR CAUSE will indicate "Failure in Protection Engines Validation" and the SCTP association will be terminated. If the association was not aborted the protected association is considered successfully established and the PROTECTED state is entered.¶
When the initiator receives the PVALID chunk, it will compare with the previous chosen Protection Engine and in case of mismatch with the one received previously in the protected association parameter in the INIT-ACK chunk, it will reply with ABORT with the ERROR CAUSE "Failure in Protection Engines Validation", otherwise the protected association is successfully established and the initiator enters the PROTECTED state.¶
If T-valid timer expires either at initiator or responder, it will generate an ABORT chunk. The ERROR handling follows what specified in Section 6.2.4.¶
In the PROTECTED state any ULP SCTP messages for any PPID MAY be exchanged in the protected SCTP association.¶
Besides the procedures for terminating an association explained in [RFC9260], the protection engine SHALL ask SCTP endpoint for terminating an association when having an internal error or by detecting a security violation, using the procedure described in Section 6.3. During the termination procedure all Control Chunks SHALL be protected except SHUTDOWN-COMPLETE. The internal design of Protection Engines and their capability is out of the scope of the current document.¶
When the Association is in PROTECTION INITILIZATION state, in-band key management shall exploit SCTP DATA chunk with the Protection Engine PPID (see Table 9) that will be sent unencrypted.¶
When the Association is in crypto chunk PROTECTED state and the SCTP assocation is in ESTABLISHED or any of the states that can be reached after ESTABLISHED state, in-band key management shall exploit SCTP DATA chunk that will be protected by the Protection Engine and encapsulated in CRYPTO chunks.¶
In-band key management shall use a dedicated Payload Protocol Identifier assigned by IANA and defined in the specific Protection Engine Specification.¶
The timer T-Valid supervises initializations that depend on how the handshake is specified for the Protection Engine and also on the characteristics of the transport network.¶
This specification recommends a default value of 30 seconds for T-valid. This value is expected to be superseded by recommendations in the Protection Engine Specification for each Protection Engine.¶
With reference to the Crypto Chunk states and the state Diagram as shown in Figure 3 of [RFC9260], the handling of Control chunks, Data chunks and Crypto chunks follows the rules defined below:¶
The diagram shown in Figure 8 describes the structure of any plain text SCTP packet being sent or received when the Crypto Chunk is not in VALIDATION or PROTECTED state.¶
The diagram shown in Figure 9 describes the structure of an SCTP packet being sent after the Crypto Chunk VALIDATION or PROTECTED state has been reached. Such packets are built with the SCTP common header. Only one CRYPTO chunk can be sent in a SCTP packet.¶
When the Crypto Chunk state machine hasn't reached the VALIDATION state, the protection enigne MAY perform protection engine key management in-band depending on how the specification for the chosen Protection Engine has been defined. In such case, the CRYPTO chunk Handler will receive plain control and DATA chunks from the SCTP chunk handler.¶
When the Crypto Chunk has reached the VALIDATION and PROTECTED state, the CRYPTO chunk handler will receive control chunks and DATA chunks from the SCTP chunk handler as a complete SCTP payload with maximum size limited by PMTU reduced by the size of the SCTP common header and the CRYPTO chunk overhead.¶
That plain payload will be sent to the protection engine in use for that specific association, the protection engine will return an encrypted payload.¶
Depending on the specification for the chosen protection engine, when forming the CRYPTO chunk header the CRYPTO chunk handler MAY set the chunk header flags (see Figure 3).¶
An SCTP packet containing an SCTP CRYPTO chunk SHALL be delivered without delay and SCTP bundling SHALL NOT be performed.¶
When the Crypto Chunk state machine hasn't reached the VALIDATION state, it MAY handle key management in-band depending on how the specification for the chosen protection engine has been defined. In such case, the CRYPTO chunk handler will receive plain control chunks and DATA chunks with Protection Engine PPID from the SCTP Header Handler. Those plain control chunks will be forwarded to SCTP chunk handler.¶
When the Crypto Chunk state machine has reached the VALIDATION or PROTECTED state, the CRYPTO chunk handler will receive CRYPTO chunks from the SCTP Header Handler. Payload from CRYPTO chunks will be forwarded to the protection engine in use for that specific association for decryption, the protection engine will return a plain SCTP Payload. The plain SCTP payload will be forwarded to SCTP Chunk Handler that will split it in separated chunks and will handle them according to [RFC9260].¶
Depending on the specification for the chosen protection engine, when receiving the CRYPTO chunk header the CRYPTO Chunk Handler MAY handle the Flags (see Figure 3) according to that specification.¶
Meta data belonging to the SCTP packet received SHALL be tied to the relevant chunks and forwarded transparently to the SCTP endpoint.¶
The SCTP Header Handler is responsible for correctness of the SCTP common header, it receives the SCTP packet from the lower transport layer, discriminates among associations and forwards the payload and relevant data to the SCTP protection engine for handling.¶
In the opposite direction it creates the SCTP common header and fills it with the relevant information for the specific association and delivers it towards the lower transport layer.¶
This document defines two new registries in the Stream Control Transmission Protocol (SCTP) Parameters group that IANA maintains. Theses registries are for the protection engine identifiers and extra cause codes for protection related errors. It also adds registry entries into several other registries in the Stream Control Transmission Protocol (SCTP) Parameters group:¶
IANA is requested to create a new registry called "CRYPTO Chunk Protection Engine Identifiers". This registry is part of the Stream Control Transmission Protocol (SCTP) Parameters grouping.¶
The purpose of this registry is to enable identification of different protection engines used by the CRYPTO chunk when performing the SCTP handshake and negotiating support. Entries in the registry requires a protection engine name, a reference to the specification for the protection engine, and a contact. Each entry will be assigned by IANA a unique 16-bit unsigned integer identifier for their protection engine. Values 0-65534 are available for assignment. Value 65535 is reserved for future extension. The proposed general form of the registry is depicted below in Table 4.¶
| ID Value | Name | Reference | Contact |
|---|---|---|---|
| 0-65534 | Available for Assignment | RFC-To-Be | |
| 65535 | Reserved | RFC-To-Be | Authors |
New entries are registered following the Specification Required policy as defined by [RFC8126].¶
IANA is requested to create a new registry called "Protection Error Cause Codes". This registry is part of the Stream Control Transmission Protocol (SCTP) Parameters grouping.¶
The purpose of this registry is to enable identification of different protection related errors when using CRYPTO chunk and a protection engine. Entries in the registry requires a Meaning, a reference to the specification defining the error, and a contact. Each entry will be assigned by IANA a unique 16-bit unsigned integer identifier for their protection engine. Values 0-65534 are available for assignment. Value 65535 is reserved for future extension. The proposed general form of the registry is depicted below in Table 5.¶
| Cause Code | Meaning | Reference | Contact |
|---|---|---|---|
| 0 | Error in the Protection Engine List | RFC-To-Be | Authors |
| 1 | Error During Protection Handshake | RFC-To-Be | Authors |
| 2 | Failure in Protection Engines Validation | RFC-To-Be | Authors |
| 3 | Timeout During KEY Handshake or Validation | RFC-To-Be | Authors |
| 4-65534 | Available for Assignment | RFC-To-Be | Authors |
| 65535 | Reserved | RFC-To-Be | Authors |
New entries are registered following the Specification Required policy as defined by [RFC8126].¶
In the Stream Control Transmission Protocol (SCTP) Parameters group's "Chunk Types" registry, IANA is requested to add the two new entries depicted below in in Table 6 with a reference to this document. The registry at time of writing was available at: https://www.iana.org/assignments/sctp-parameters/sctp-parameters.xhtml#sctp-parameters-1¶
| ID Value | Chunk Type | Reference |
|---|---|---|
| TBA6 | Crypto Chunk (CRYPTO) | RFC-To-Be |
| TBA7 | Protected Association Parameter Validation (PVALID) | RFC-To-Be |
In the Stream Control Transmission Protocol (SCTP) Parameters group's "Chunk Parameter Types" registry, IANA is requested to add the new entry depicted below in in Table 7 with a reference to this document. The registry at time of writing was available at: https://www.iana.org/assignments/sctp-parameters/sctp-parameters.xhtml#sctp-parameters-2¶
| ID Value | Chunk Parameter Type | Reference |
|---|---|---|
| TBA8 | Protected Association | RFC-To-Be |
In the Stream Control Transmission Protocol (SCTP) Parameters group's "Error Cause Codes" registry, IANA is requested to add the new entry depicted below in in Table 8 with a reference to this document. The registry at time of writing was available at: https://www.iana.org/assignments/sctp-parameters/sctp-parameters.xhtml#sctp-parameters-24¶
| ID Value | Error Cause Codes | Reference |
|---|---|---|
| TBA9 | Protection Engine Error | RFC-To-Be |
In the Stream Control Transmission Protocol (SCTP) Parameters group's "Payload Protocol Identifiers" registry, IANA is requested to add the new entry depicted below in in Table 9 with a reference to this document. The registry at time of writing was available at: https://www.iana.org/assignments/sctp-parameters/sctp-parameters.xhtml#sctp-parameters-25¶
| ID Value | SCTP Payload Protocol Identifier | Reference |
|---|---|---|
| TBA10 | Protection Engine Protocol Identifier | RFC-To-Be |
All the security and privacy considerations of the security protocol used as the protection engine applies.¶
Using a security protocol in the SCTP CRYPTO chunk might lower the privacy properties of the security protocol as the SCTP Verification Tag is an unique identifier for the association.¶
The CRYPTO chunk provides a mechanism for preventing downgrade attacks that detects downgrading attempts between protection engines and terminates the association. The chosen protection engine is the same as if the peers had been communicating in the absence of an attacker.¶
The protection engine initial handshake is verified before the Crypto Chunk is considered protected, thus no user data are sent before validation.¶
The downgrade protection is only as strong as the weakest of the supported protection engines as an active attacker can trick the endpoints to negotiate the weakest protection engine and then modify the weakly protected CRYPTO chunks to deceive the endpoints that the negotiation of the protection engines is validated. This is similar to the downgrade protection in TLS 1.3 specified in Section 4.1.3. of [RFC8446] where downgrade protection is not provided when TLS 1.2 with static RSA is used. It is RECOMMENDED to only support a limited set of strongly profiled protection engines.¶
This section specifies what is to be specified in the description of a protection engine.¶
The authors thank Michael Tüxen for his invaluable comments helping to cope with Association Restart, ASCONF handling and restructuring the Protection Engine architecture.¶