Internet-Draft | TLS/DTLS 1.3 IoT Profiles | March 2023 |
Tschofenig & Fossati | Expires 14 September 2023 | [Page] |
This document is a companion to RFC 7925 and defines TLS/DTLS 1.3 profiles for Internet of Things devices. It also updates RFC 7925 with regards to the X.509 certificate profile.¶
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 14 September 2023.¶
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 profile of DTLS 1.3 [DTLS13] and TLS 1.3 [RFC8446] that offers communication security services for IoT applications and is reasonably implementable on many constrained devices. Profile thereby means that available configuration options and protocol extensions are utilized to best support the IoT environment.¶
For IoT profiles using TLS/DTLS 1.2 please consult [RFC7925]. This document re-uses the communication pattern defined in [RFC7925] and makes IoT-domain specific recommendations for version 1.3 (where necessary).¶
TLS 1.3 has been re-designed and several previously defined extensions are not applicable to the new version of TLS/DTLS anymore. This clean-up also simplifies this document. Furthermore, many outdated ciphersuites have been omitted from the TLS/DTLS 1.3 specification.¶
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.¶
In accordance with the recommendations in [RFC7925], a compliant implementation MUST implement TLS_AES_128_CCM_8_SHA256. It SHOULD implement TLS_CHACHA20_POLY1305_SHA256.¶
Pre-shared key based authentication is integrated into the main TLS/DTLS 1.3 specification and has been harmonized with session resumption.¶
A compliant implementation supporting authentication based on certificates and raw public keys MUST support digital signatures with ecdsa_secp256r1_sha256. A compliant implementation MUST support the key exchange with secp256r1 (NIST P-256) and SHOULD support key exchange with X25519.¶
A plain PSK-based TLS/DTLS client or server MUST implement the following extensions:¶
For use of external pre-shared keys [RFC9258] makes the following recommendation:¶
Applications SHOULD provision separate PSKs for (D)TLS 1.3 and prior versions.¶
Where possible, the importer interface defined in [RFC9258] MUST be used for external PSKs. This ensures that external PSKs used in (D)TLS 1.3 are bound to a specific key derivation function (KDF) and hash function.¶
The SNI extension is discussed in this document and the justification for implementing and using the ALPN extension can be found in [RFC9325].¶
For TLS/DTLS clients and servers implementing raw public keys and/or certificates the guidance for mandatory-to-implement extensions described in Section 9.2 of [RFC8446] MUST be followed.¶
TLS 1.3 simplified the Alert protocol but the underlying challenge in an embedded context remains unchanged, namely what should an IoT device do when it encounters an error situation. The classical approach used in a desktop environment where the user is prompted is often not applicable with unattended devices. Hence, it is more important for a developer to find out from which error cases a device can recover from.¶
TLS 1.3 has built-in support for session resumption by utilizing PSK-based credentials established in an earlier exchange.¶
TLS 1.3 does not have support for compression of application data traffic, as offered by previous versions of TLS. Applications are therefore responsible for transmitting payloads that are either compressed or use a more efficient encoding otherwise.¶
With regards to the handshake itself, various strategies have been applied to reduce the size of the exchanged payloads. TLS and DTLS 1.3 use less overhead, depending on the type of key confirmations, when compared to previous versions of the protocol. Additionally, the work on Compact TLS (cTLS) [I-D.ietf-tls-ctls] has taken compression of the handshake a step further by utilizing out-of-band knowledge between the communication parties to reduce the amount of data to be transmitted at each individual handshake, among applying other techniques.¶
TLS 1.3 allows the use of PFS with all ciphersuites since the support for it is negotiated independently.¶
The recommendation in Section 11 of [RFC7925] is applicable. In particular this document RECOMMENDED to use an initial timer value of 9 seconds with exponential back off up to no less then 60 seconds.¶
The discussion in Section 12 of [RFC7925] is applicable with one exception: the ClientHello and the ServerHello messages in TLS 1.3 do not contain gmt_unix_time component anymore.¶
This specification mandates the implementation of the Server Name Indication (SNI) extension. Where privacy requirements require it, the Encrypted Client Hello extension [I-D.ietf-tls-esni] prevents an on-path attacker to determine the domain name the client is trying to connect to.¶
Note: To avoid leaking DNS lookups from network inspection altogether further protocols are needed, including DoH [RFC8484] and DPRIVE [RFC7858] [RFC8094]. Since the Encrypted Client Hello extension requires use of Hybrid Public Key Encryption (HPKE) [I-D.irtf-cfrg-hpke] and additional protocols require further protocol exchanges and cryptographic operations, there is a certain overhead associated with this privacy feature.¶
The Maximum Fragment Length Negotiation (MFL) extension has been superseded by the Record Size Limit (RSL) extension [RFC8449]. Implementations in compliance with this specification MUST implement the RSL extension and SHOULD use it to indicate their RAM limitations.¶
The recommendations in Section 19 of [RFC7925] are applicable.¶
The recommendations in Section 20 of [RFC7925] are applicable.¶
Appendix E.5 of [TLS13] establishes that:¶
Application protocols MUST NOT use 0-RTT data without a profile that defines its use. That profile needs to identify which messages or interactions are safe to use with 0-RTT and how to handle the situation when the server rejects 0-RTT and falls back to 1-RTT.¶
At the time of writing, no such profile has been defined for CoAP [CoAP]. Therefore, 0-RTT MUST NOT be used by CoAP applications.¶
This section contains updates and clarifications to the certificate profile defined in [RFC7925]. The content of Table 1 of [RFC7925] has been split by certificate "type" in order to clarify exactly what requirements and recommendations apply to which entity in the PKI hierarchy.¶
CAs SHALL generate non-sequential certificate serial numbers greater than zero (0) containing at least 64 bits of output from a CSPRNG (cryptographically secure pseudo-random number generator).¶
No maximum validity period is mandated. Validity values are expressed in notBefore and notAfter fields, as described in Section 4.1.2.5 of [RFC5280]. In particular, values MUST be expressed in Greenwich Mean Time (Zulu) and MUST include seconds even where the number of seconds is zero.¶
Note that the validity period is defined as the period of time from notBefore through notAfter, inclusive. This means that a hypothetical certificate with a notBefore date of 9 June 2021 at 03:42:01 and a notAfter date of 7 September 2021 at 03:42:01 becomes valid at the beginning of the :01 second, and only becomes invalid at the :02 second, a period that is 90 days plus 1 second. So for a 90-day, notAfter must actually be 03:42:00.¶
In many cases it is necessary to indicate that a certificate does not expire. This is likely to be the case for manufacturer-provisioned certificates. RFC 5280 provides a simple solution to convey the fact that a certificate has no well-defined expiration date by setting the notAfter to the GeneralizedTime value of 99991231235959Z.¶
Some devices might not have a reliable source of time and for those devices it is also advisable to use certificates with no expiration date and to let a device management solution manage the lifetime of all the certificates used by the device. While this approach does not utilize certificates to its widest extent, it is a solution that extends the capabilities offered by a raw public key approach.¶
The SubjectPublicKeyInfo structure indicates the algorithm and any associated parameters for the ECC public key. This profile uses the id-ecPublicKey algorithm identifier for ECDSA signature keys, as defined and specified in [RFC5480].¶
*
) characters. subjectAltName MUST NOT contain multiple
names.¶
The requirement in Section 4.4.2 of [RFC7925] to only use EUI-64 for client certificates is lifted.¶
If the EUI-64 format is used to identify the subject of a client certificate,
it MUST be encoded in a subjectAltName of type DNS-ID as a string of the form
HH-HH-HH-HH-HH-HH-HH-HH
where 'H' is one of the symbols '0'-'9' or 'A'-'F'.¶
The considerations in Section 4.4.3 of [RFC7925] hold.¶
Since the publication of RFC 7925 the need for firmware update mechanisms has been reinforced and the work on standardizing a secure and interoperable firmware update mechanism has made substantial progress, see [RFC9019]. RFC 7925 recommends to use a software / firmware update mechanism to provision devices with new trust anchors.¶
The use of device management protocols for IoT devices, which often include an onboarding or bootstrapping mechanism, has also seen considerable uptake in deployed devices and these protocols, some of which are standardized, allow provision of certificates on a regular basis. This enables a deployment model where IoT device utilize end-entity certificates with shorter lifetime making certificate revocation protocols, like OCSP and CRLs, less relevant.¶
Hence, instead of performing certificate revocation checks on the IoT device itself this specification recommends to delegate this task to the IoT device operator and to take the necessary action to allow IoT devices to remain operational.¶
In a public key-based key exchange, certificates and public keys are a major contributor to the size of the overall handshake. For example, in a regular TLS 1.3 handshake with minimal ECC certificates and no subordinate CA utilizing the secp256r1 curve with mutual authentication, around 40% of the entire handshake payload is consumed by the two exchanged certificates.¶
Hence, it is not surprising that there is a strong desire to reduce the size of certificates and certificate chains. This has lead to various standardization efforts. Here is a brief summary of what options an implementer has to reduce the bandwidth requirements of a public key-based key exchange:¶
The use of certificate handles, as introduced in cTLS [I-D.ietf-tls-ctls], is a form of caching or compressing certificates as well.¶
Whether to utilize any of the above extensions or a combination of them depends on the anticipated deployment environment, the availability of code, and the constraints imposed by already deployed infrastructure (e.g., CA infrastructure, tool support).¶
Section 4.5.3 of [DTLS13] flags AES-CCM with 8-octet authentication tags (CCM_8) as unsuitable for general use with DTLS. In fact, due to its low integrity limits (i.e., a high sensitivity to forgeries), endpoints that negotiate ciphersuites based on such AEAD are susceptible to a trivial DoS. (See also Section 5.3 and 5.4 of [I-D.irtf-cfrg-aead-limits] for further discussion on this topic, as well as references to the analysis supporting these conclusions.)¶
Specifically, [DTLS13] warns that:¶
"TLS_AES_128_CCM_8_SHA256 MUST NOT be used in DTLS without additional safeguards against forgery. Implementations MUST set usage limits for AEAD_AES_128_CCM_8 based on an understanding of any additional forgery protections that are used."¶
Since all the ciphersuites mandated by [RFC7925] and [CoAP] are based on CCM_8, there is no stand-by ciphersuite to use for applications that want to avoid the security and availability risks associated with CCM_8 while retaining interoperability with the rest of the ecosystem.¶
In order to ameliorate the situation, this document RECOMMENDS that implementations support the following two ciphersuites:¶
and offer them as their first choice. These ciphersuites provide confidentiality and integrity limits that are considered acceptable in the most general settings. For the details on the exact bounds of both ciphersuites see Section 4.5.3 of [DTLS13]. Note that the GCM-based ciphersuite offers superior interoperability with cloud services at the cost of a slight increase in the wire and peak RAM footprints.¶
When the GCM-based ciphersuite is used with TLS 1.2, the recommendations in Section 6.2.1 of [RFC9325] related to deterministic nonce generation apply. In addition, the integrity limits on key usage detailed in Section 4.4 of [RFC9325] also apply.¶
A number of passive side-channel attacks as well as active fault-injection attacks (e.g., [Ambrose2017]) have been demonstrated that allow a malicious third party to gain information about the signing key if a fully deterministic signature scheme (e.g., [RFC6979] ECDSA or EdDSA [RFC8032]) is used.¶
Most of these attacks assume physical access to the device and are therefore especially relevant to smart cards as well as IoT deployments with poor or non-existent physical security.¶
In this security model, it is recommended to combine both randomness and determinism, for example, as described in [I-D.mattsson-cfrg-det-sigs-with-noise].¶
A list of open issues can be found at https://github.com/thomas-fossati/draft-tls13-iot/issues¶
This entire document is about security.¶
This document makes no requests to IANA.¶
We would like to thank Ben Kaduk and John Mattsson.¶