Internet Engineering Task Force (IETF) D. McGrew Request for Comments: 5764 Cisco Systems Category: Standards Track E. Rescorla ISSN: 2070-1721 RTFM, Inc. May 2010
Datagram Transport Layer Security (DTLS) Extension to Establish Keys for the Secure Real-time Transport Protocol (SRTP)
Abstract
This document describes a Datagram Transport Layer Security (DTLS) extension to establish keys for Secure RTP (SRTP) and Secure RTP Control Protocol (SRTCP) flows. DTLS keying happens on the media path, independent of any out-of-band signalling channel present.
Status of This Memo
This is an Internet Standards Track document.
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). Further information on Internet Standards is available in Section 2 of RFC 5741.
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc5764.
Copyright Notice
Copyright (c) 2010 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 (http://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 Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
McGrew & Rescorla Standards Track [Page 1]
RFC 5764 SRTP Extension for DTLS May 2010
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.
The Secure RTP (SRTP) profile [RFC3711] can provide confidentiality, message authentication, and replay protection to RTP data and RTP Control (RTCP) traffic. SRTP does not provide key management functionality, but instead depends on external key management to exchange secret master keys, and to negotiate the algorithms and parameters for use with those keys.
Datagram Transport Layer Security (DTLS) [RFC4347] is a channel security protocol that offers integrated key management, parameter negotiation, and secure data transfer. Because DTLS data transfer protocol is generic, it is less highly optimized for use with RTP than is SRTP, which has been specifically tuned for that purpose.
This document describes DTLS-SRTP, a SRTP extension for DTLS that combines the performance and encryption flexibility benefits of SRTP with the flexibility and convenience of DTLS-integrated key and association management. DTLS-SRTP can be viewed in two equivalent ways: as a new key management method for SRTP, and a new RTP-specific data format for DTLS.
The key points of DTLS-SRTP are that:
o application data is protected using SRTP,
o the DTLS handshake is used to establish keying material, algorithms, and parameters for SRTP,
o a DTLS extension is used to negotiate SRTP algorithms, and
o other DTLS record-layer content types are protected using the ordinary DTLS record format.
The remainder of this memo is structured as follows. Section 2 describes conventions used to indicate normative requirements. Section 3 provides an overview of DTLS-SRTP operation. Section 4 specifies the DTLS extensions, while Section 5 discusses how RTP and RTCP are transported over a DTLS-SRTP channel. Section 6 describes use with multi-party sessions. Section 7 and Section 9 describe Security and IANA considerations.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
DTLS-SRTP is defined for point-to-point media sessions, in which there are exactly two participants. Each DTLS-SRTP session contains a single DTLS association (called a "connection" in TLS jargon), and either two SRTP contexts (if media traffic is flowing in both directions on the same host/port quartet) or one SRTP context (if media traffic is only flowing in one direction). All SRTP traffic flowing over that pair in a given direction uses a single SRTP context. A single DTLS-SRTP session only protects data carried over a single UDP source and destination port pair.
The general pattern of DTLS-SRTP is as follows. For each RTP or RTCP flow the peers do a DTLS handshake on the same source and destination port pair to establish a DTLS association. Which side is the DTLS client and which side is the DTLS server must be established via some out-of-band mechanism such as SDP. The keying material from that handshake is fed into the SRTP stack. Once that association is established, RTP packets are protected (becoming SRTP) using that keying material.
RTP and RTCP traffic is usually sent on two separate UDP ports. When symmetric RTP [RFC4961] is used, two bidirectional DTLS-SRTP sessions are needed, one for the RTP port, one for the RTCP port. When RTP flows are not symmetric, four unidirectional DTLS-SRTP sessions are needed (for inbound and outbound RTP, and inbound and outbound RTCP).
Symmetric RTP [RFC4961] is the case in which there are two RTP sessions that have their source and destination ports and addresses reversed, in a manner similar to the way that a TCP connection uses its ports. Each participant has an inbound RTP session and an outbound RTP session. When symmetric RTP is used, a single DTLS-SRTP session can protect both of the RTP sessions. It is RECOMMENDED that symmetric RTP be used with DTLS-SRTP.
RTP and RTCP traffic MAY be multiplexed on a single UDP port [RFC5761]. In this case, both RTP and RTCP packets may be sent over the same DTLS-SRTP session, halving the number of DTLS-SRTP sessions needed. This improves the cryptographic performance of DTLS, but may cause problems when RTCP and RTP are subject to different network treatment (e.g., for bandwidth reservation or scheduling reasons).
Between a single pair of participants, there may be multiple media sessions. There MUST be a separate DTLS-SRTP session for each distinct pair of source and destination ports used by a media session (though the sessions can share a single DTLS session and hence amortize the initial public key handshake!).
McGrew & Rescorla Standards Track [Page 4]
RFC 5764 SRTP Extension for DTLS May 2010
A DTLS-SRTP session may be indicated by an external signaling protocol like SIP. When the signaling exchange is integrity- protected (e.g., when SIP Identity protection via digital signatures is used), DTLS-SRTP can leverage this integrity guarantee to provide complete security of the media stream. A description of how to indicate DTLS-SRTP sessions in SIP and SDP [RFC4566], and how to authenticate the endpoints using fingerprints can be found in [RFC5763].
In a naive implementation, when there are multiple media sessions, there is a new DTLS session establishment (complete with public key cryptography) for each media channel. For example, a videophone may be sending both an audio stream and a video stream, each of which would use a separate DTLS session establishment exchange, which would proceed in parallel. As an optimization, the DTLS-SRTP implementation SHOULD use the following strategy: a single DTLS association is established, and all other DTLS associations wait until that connection is established before proceeding with their handshakes. This strategy allows the later sessions to use DTLS session resumption, which allows the amortization of the expensive public key cryptography operations over multiple DTLS handshakes.
The SRTP keys used to protect packets originated by the client are distinct from the SRTP keys used to protect packets originated by the server. All of the RTP sources originating on the client for the same channel use the same SRTP keys, and similarly, all of the RTP sources originating on the server for the same channel use the same SRTP keys. The SRTP implementation MUST ensure that all of the synchronization source (SSRC) values for all of the RTP sources originating from the same device over the same channel are distinct, in order to avoid the "two-time pad" problem (as described in Section 9.1 of RFC 3711). Note that this is not an issue for separate media streams (on different host/port quartets) that use independent keying material even if an SSRC collision occurs.
In order to negotiate the use of SRTP data protection, clients include an extension of type "use_srtp" in the DTLS extended client hello. This extension MUST only be used when the data being transported is RTP or RTCP [RFC3550]. The "extension_data" field of this extension contains the list of acceptable SRTP protection profiles, as indicated below.
McGrew & Rescorla Standards Track [Page 5]
RFC 5764 SRTP Extension for DTLS May 2010
Servers that receive an extended hello containing a "use_srtp" extension can agree to use SRTP by including an extension of type "use_srtp", with the chosen protection profile in the extended server hello. This process is shown below.
Note that '*' indicates messages that are not always sent in DTLS. The CertificateRequest, client and server Certificates, and CertificateVerify will be sent in DTLS-SRTP.
Once the "use_srtp" extension is negotiated, the RTP or RTCP application data is protected solely using SRTP. Application data is never sent in DTLS record-layer "application_data" packets. Rather, complete RTP or RTCP packets are passed to the DTLS stack, which passes them to the SRTP stack, which protects them appropriately. Note that if RTP/RTCP multiplexing [RFC5761] is in use, this means that RTP and RTCP packets may both be passed to the DTLS stack. Because the DTLS layer does not process the packets, it does not need to distinguish them. The SRTP stack can use the procedures of [RFC5761] to distinguish RTP from RTCP.
When the "use_srtp" extension is in effect, implementations must not place more than one application data "record" per datagram. (This is only meaningful from the perspective of DTLS because SRTP is inherently oriented towards one payload per packet, but this is stated purely for clarification.)
Data other than RTP/RTCP (i.e., TLS control messages) MUST use ordinary DTLS framing and MUST be placed in separate datagrams from SRTP data.
McGrew & Rescorla Standards Track [Page 6]
RFC 5764 SRTP Extension for DTLS May 2010
A DTLS-SRTP handshake establishes one or more SRTP crypto contexts; however, they all have the same SRTP Protection Profile and Master Key Identifier (MKI), if any. MKIs are used solely to distinguish the keying material and protection profiles between distinct handshakes, for instance, due to rekeying. When an MKI is established in a DTLS-SRTP session, it MUST apply for all of the SSRCs within that session -- though a single endpoint may negotiate multiple DTLS-SRTP sessions due, for instance, to forking. (Note that RFC 3711 allows packets within the same session but with different SSRCs to use MKIs differently; in contrast, DTLS-SRTP requires that MKIs and the keys that they are associated with have the same meaning and are uniform across the entire SRTP session.)
The SRTPProtectionProfiles list indicates the SRTP protection profiles that the client is willing to support, listed in descending order of preference. The srtp_mki value contains the SRTP Master Key Identifier (MKI) value (if any) that the client will use for his SRTP packets. If this field is of zero length, then no MKI will be used.
Note: for those unfamiliar with TLS syntax, "srtp_mki<0..255>" indicates a variable-length value with a length between 0 and 255 (inclusive). Thus, the MKI may be up to 255 bytes long.
If the server is willing to accept the use_srtp extension, it MUST respond with its own "use_srtp" extension in the ExtendedServerHello. The extension_data field MUST contain a UseSRTPData value with a single SRTPProtectionProfile value that the server has chosen for use with this connection. The server MUST NOT select a value that the client has not offered. If there is no shared profile, the server SHOULD NOT return the use_srtp extension at which point the connection falls back to the negotiated DTLS cipher suite. If that is not acceptable, the server SHOULD return an appropriate DTLS alert.
A DTLS-SRTP SRTP Protection Profile defines the parameters and options that are in effect for the SRTP processing. This document defines the following SRTP protection profiles.
The following list indicates the SRTP transform parameters for each protection profile. The parameters cipher_key_length, cipher_salt_length, auth_key_length, and auth_tag_length express the number of bits in the values to which they refer. The maximum_lifetime parameter indicates the maximum number of packets that can be protected with each single set of keys when the parameter profile is in use. All of these parameters apply to both RTP and RTCP, unless the RTCP parameters are separately specified.
All of the crypto algorithms in these profiles are from [RFC3711].
With all of these SRTP Parameter profiles, the following SRTP options are in effect:
o The TLS PseudoRandom Function (PRF) is used to generate keys to feed into the SRTP Key Derivation Function (KDF). When DTLS 1.2 [DTLS1.2] is in use, the PRF is the one associated with the cipher suite. Note that this specification is compatible with DTLS 1.0 or DTLS 1.2
o The Key Derivation Rate (KDR) is equal to zero. Thus, keys are not re-derived based on the SRTP sequence number.
o The key derivation procedures from Section 4.3 with the AES-CM PRF from RFC 3711 are used.
o For all other parameters (in particular, SRTP replay window size and FEC order), the default values are used.
If values other than the defaults for these parameters are required, they can be enabled by writing a separate specification specifying SDP syntax to signal them.
Applications using DTLS-SRTP SHOULD coordinate the SRTP Protection Profiles between the DTLS-SRTP session that protects an RTP flow and the DTLS-SRTP session that protects the associated RTCP flow (in those cases in which the RTP and RTCP are not multiplexed over a common port). In particular, identical ciphers SHOULD be used.
New SRTPProtectionProfile values must be defined according to the "Specification Required" policy as defined by RFC 5226 [RFC5226]. See Section 9 for IANA Considerations.
The srtp_mki value MAY be used to indicate the capability and desire to use the SRTP Master Key Identifier (MKI) field in the SRTP and SRTCP packets. The MKI field indicates to an SRTP receiver which key was used to protect the packet that contains that field. The
McGrew & Rescorla Standards Track [Page 9]
RFC 5764 SRTP Extension for DTLS May 2010
srtp_mki field contains the value of the SRTP MKI which is associated with the SRTP master keys derived from this handshake. Each SRTP session MUST have exactly one master key that is used to protect packets at any given time. The client MUST choose the MKI value so that it is distinct from the last MKI value that was used, and it SHOULD make these values unique for the duration of the TLS session.
Upon receipt of a "use_srtp" extension containing a "srtp_mki" field, the server MUST either (assuming it accepts the extension at all):
1. include a matching "srtp_mki" value in its "use_srtp" extension to indicate that it will make use of the MKI, or 2. return an empty "srtp_mki" value to indicate that it cannot make use of the MKI.
If the client detects a nonzero-length MKI in the server's response that is different than the one the client offered, then the client MUST abort the handshake and SHOULD send an invalid_parameter alert. If the client and server agree on an MKI, all SRTP packets protected under the new security parameters MUST contain that MKI.
Note that any given DTLS-SRTP session only has a single active MKI (if any). Thus, at any given time, a set of endpoints will generally only be using one MKI (the major exception is during rehandshakes).
When SRTP mode is in effect, different keys are used for ordinary DTLS record protection and SRTP packet protection. These keys are generated using a TLS exporter [RFC5705] to generate
2 * (SRTPSecurityParams.master_key_len + SRTPSecurityParams.master_salt_len) bytes of data
which are assigned as shown below. The per-association context value is empty.
The exporter label for this usage is "EXTRACTOR-dtls_srtp". (The "EXTRACTOR" prefix is for historical compatibility.)
The four keying material values (the master key and master salt for each direction) are provided as inputs to the SRTP key derivation mechanism, as shown in Figure 1 and detailed below. By default, the
McGrew & Rescorla Standards Track [Page 10]
RFC 5764 SRTP Extension for DTLS May 2010
mechanism defined in Section 4.3 of [RFC3711] is used, unless another key derivation mechanism is specified as part of an SRTP Protection Profile.
The client_write_SRTP_master_key and client_write_SRTP_master_salt are provided to one invocation of the SRTP key derivation function, to generate the SRTP keys used to encrypt and authenticate packets sent by the client. The server MUST only use these keys to decrypt and to check the authenticity of inbound packets.
The server_write_SRTP_master_key and server_write_SRTP_master_salt are provided to one invocation of the SRTP key derivation function, to generate the SRTP keys used to encrypt and authenticate packets sent by the server. The client MUST only use these keys to decrypt and to check the authenticity of inbound packets.
When both RTCP and RTP use the same source and destination ports, then both the SRTP and SRTCP keys are needed. Otherwise, there are two DTLS-SRTP sessions, one of which protects the RTP packets and one of which protects the RTCP packets; each DTLS-SRTP session protects the part of an SRTP session that passes over a single source/ destination transport address pair, as shown in Figure 2, independent of which SSRCs are used on that pair. When a DTLS-SRTP session is protecting RTP, the SRTCP keys derived from the DTLS handshake are not needed and are discarded. When a DTLS-SRTP session is protecting RTCP, the SRTP keys derived from the DTLS handshake are not needed and are discarded.
Because of the possibility of packet reordering, DTLS-SRTP implementations SHOULD store multiple SRTP keys sets during a rekey in order to avoid the need for receivers to drop packets for which they lack a key.
The maximum_lifetime parameter in the SRTP protection profile indicates the maximum number of packets that can be protected with each single encryption and authentication key. (Note that, since RTP and RTCP are protected with independent keys, those protocols are counted separately for the purposes of determining when a key has reached the end of its lifetime.) Each profile defines its own limit. When this limit is reached, a new DTLS session SHOULD be used to establish replacement keys, and SRTP implementations MUST NOT use the existing keys for the processing of either outbound or inbound traffic.
Once the DTLS handshake has completed, the peers can send RTP or RTCP over the newly created channel. We describe the transmission process first followed by the reception process.
Within each RTP session, SRTP processing MUST NOT take place before the DTLS handshake completes.
DTLS and TLS define a number of record content types. In ordinary TLS/DTLS, all data is protected using the same record encoding and mechanisms. When the mechanism described in this document is in effect, this is modified so that data written by upper-level protocol clients of DTLS is assumed to be RTP/RTP and is encrypted using SRTP rather than the standard TLS record encoding.
When a user of DTLS wishes to send an RTP packet in SRTP mode, it delivers it to the DTLS implementation as an ordinary application data write (e.g., SSL_write()). The DTLS implementation then invokes the processing described in RFC 3711, Sections 3 and 4. The resulting SRTP packet is then sent directly on the wire as a single datagram with no DTLS framing. This provides an encapsulation of the data that conforms to and interoperates with SRTP. Note that the RTP sequence number rather than the DTLS sequence number is used for these packets.
When DTLS-SRTP is used to protect an RTP session, the RTP receiver needs to demultiplex packets that are arriving on the RTP port. Arriving packets may be of types RTP, DTLS, or STUN [RFC5389]. If these are the only types of packets present, the type of a packet can be determined by looking at its first byte.
The process for demultiplexing a packet is as follows. The receiver looks at the first byte of the packet. If the value of this byte is 0 or 1, then the packet is STUN. If the value is in between 128 and 191 (inclusive), then the packet is RTP (or RTCP, if both RTCP and RTP are being multiplexed over the same destination port). If the value is between 20 and 63 (inclusive), the packet is DTLS. This process is summarized in Figure 3.
McGrew & Rescorla Standards Track [Page 13]
RFC 5764 SRTP Extension for DTLS May 2010
+----------------+ | 127 < B < 192 -+--> forward to RTP | | packet --> | 19 < B < 64 -+--> forward to DTLS | | | B < 2 -+--> forward to STUN +----------------+
Figure 3: The DTLS-SRTP receiver's packet demultiplexing algorithm. Here the field B denotes the leading byte of the packet.
If other packet types are to be multiplexed as well, implementors and/or designers SHOULD ensure that they can be demultiplexed from these three packet types.
In some cases, there will be multiple DTLS-SRTP associations for a given SRTP endpoint. For instance, if Alice makes a call that is SIP forked to both Bob and Charlie, she will use the same local host/port pair for both of them, as shown in Figure 4, where XXX and YYY represent different DTLS-SRTP associations. (The SSRCs shown are the ones for data flowing to Alice.)
Bob (192.0.2.1:6666) / / / SSRC=1 / DTLS-SRTP=XXX / v Alice (192.0.2.0:5555) ^ \ \ SSRC=2 \ DTLS-SRTP=YYY \ \ Charlie (192.0.2.2:6666)
Figure 4: RTP sessions with SIP forking.
Because DTLS operates on the host/port quartet, the DTLS association will still complete correctly, with the foreign host/port pair being used, to distinguish the associations. However, in RTP the source host/port is not used and sessions are identified by the destination host/port and the SSRC. Thus, some mechanism is needed to determine which SSRCs correspond to which DTLS associations. The following method SHOULD be used.
McGrew & Rescorla Standards Track [Page 14]
RFC 5764 SRTP Extension for DTLS May 2010
For each local host/port pair, the DTLS-SRTP implementation maintains a table listing all the SSRCs it knows about and the DTLS-SRTP associations they correspond to. Initially, this table is empty. When an SRTP packet is received for a given RTP endpoint (destination IP/port pair), the following procedure is used:
1. If the SSRC is already known for that endpoint, then the corresponding DTLS-SRTP association and its keying material is used to decrypt and verify the packet. 2. If the SSRC is not known, then the receiver tries to decrypt it with the keying material corresponding to each DTLS-SRTP association for that endpoint. 3. If the decryption and verification succeeds (the authentication tag verifies), then an entry is placed in the table mapping the SSRC to that association. 4. If the decryption and verification fails, then the packet is silently discarded. 5. When a DTLS-SRTP association is closed (for instance, because the fork is abandoned), its entries MUST be removed from the mapping table.
The average cost of this algorithm for a single SSRC is the decryption and verification time of a single packet times the number of valid DTLS-SRTP associations corresponding to a single receiving port on the host. In practice, this means the number of forks; so in the case shown in Figure 4, that would be two. This cost is only incurred once for any given SSRC, since afterwards that SSRC is placed in the map table and looked up immediately. As with normal RTP, this algorithm allows new SSRCs to be introduced by the source at any time. They will automatically be mapped to the correct DTLS association.
Note that this algorithm explicitly allows multiple SSRCs to be sent from the same address/port pair. One way in which this can happen is an RTP translator. This algorithm will automatically assign the SSRCs to the correct associations. Note that because the SRTP packets are cryptographically protected, such a translator must either share keying material with one endpoint or refrain from modifying the packets in a way which would cause the integrity check to fail. This is a general property of SRTP and is not specific to DTLS-SRTP.
There are two error cases that should be considered. First, if an SSRC collision occurs, then only the packets from the first source will be processed. When the packets from the second source arrive, the DTLS association with the first source will be used for decryption and verification, which will fail, and the packet will be discarded. This is consistent with [RFC3550], which permits the
McGrew & Rescorla Standards Track [Page 15]
RFC 5764 SRTP Extension for DTLS May 2010
receiver to keep the packets from one source and discard those from the other. Of course the RFC 3550 SSRC collision detection and handling procedures MUST also be followed.
Second, there may be cases where a malfunctioning source is sending corrupt packets that cannot be decrypted and verified. In this case, the SSRC will never be entered into the mapping table because the decryption and verification always fails. Receivers MAY keep records of unmapped SSRCs that consistently fail decryption and verification and abandon attempts to process them once they reach some limit. That limit MUST be large enough to account for the effects of transmission errors. Entries MUST be pruned from this table when the relevant SRTP endpoint is deleted (e.g., the call ends) and SHOULD time out faster than that (we do not offer a hard recommendation but 10 to 30 seconds seems appropriate) in order to allow for the possibility that the peer implementation has been corrected.
Rekeying in DTLS is accomplished by performing a new handshake over the existing DTLS channel. That is, the handshake messages are protected by the existing DTLS cipher suite. This handshake can be performed in parallel with data transport, so no interruption of the data flow is required. Once the handshake is finished, the newly derived set of keys is used to protect all outbound packets, both DTLS and SRTP.
Because of packet reordering, packets protected by the previous set of keys can appear on the wire after the handshake has completed. To compensate for this fact, receivers SHOULD maintain both sets of keys for some time in order to be able to decrypt and verify older packets. The keys should be maintained for the duration of the maximum segment lifetime (MSL).
If an MKI is used, then the receiver should use the corresponding set of keys to process an incoming packet. If no matching MKI is present, the packet MUST be rejected. Otherwise, when a packet arrives after the handshake completed, a receiver SHOULD use the newly derived set of keys to process that packet unless there is an MKI. (If the packet was protected with the older set of keys, this fact will become apparent to the receiver as an authentication failure will occur.) If the authentication check on the packet fails and no MKI is being used, then the receiver MAY process the packet with the older set of keys. If that authentication check indicates that the packet is valid, the packet should be accepted; otherwise, the packet MUST be discarded and rejected.
McGrew & Rescorla Standards Track [Page 16]
RFC 5764 SRTP Extension for DTLS May 2010
Receivers MAY use the SRTP packet sequence number to aid in the selection of keys. After a packet has been received and authenticated with the new key set, any packets with sequence numbers that are greater will also have been protected with the new key set.
Since DTLS is a point-to-point protocol, DTLS-SRTP is intended only to protect unicast RTP sessions. This does not preclude its use with RTP mixers. For example, a conference bridge may use DTLS-SRTP to secure the communication to and from each of the participants in a conference. However, because each flow between an endpoint and a mixer has its own key, the mixer has to decrypt and then reencrypt the traffic for each recipient.
A future specification may describe methods for sharing a single key between multiple DTLS-SRTP associations thus allowing conferencing systems to avoid the decrypt/reencrypt stage. However, any system in which the media is modified (e.g., for level balancing or transcoding) will generally need to be performed on the plaintext and will certainly break the authentication tag, and therefore will require a decrypt/reencrypt stage.
One concern here is that attackers might be able to implement a bid- down attack forcing the peers to use ordinary DTLS rather than SRTP. However, because the negotiation of this extension is performed in the DTLS handshake, it is protected by the Finished messages. Therefore, any bid-down attack is automatically detected, which reduces this to a denial-of-service attack -- which can be mounted by any attacker who can control the channel.
Because two different framing formats are used, there is concern that an attacker could convince the receiver to treat an SRTP-framed RTP packet as a DTLS record (e.g., a handshake message) or vice versa. This attack is prevented by using different keys for Message Authentication Code (MAC) verification for each type of data. Therefore, this type of attack reduces to being able to forge a packet with a valid MAC, which violates a basic security invariant of both DTLS and SRTP.
McGrew & Rescorla Standards Track [Page 17]
RFC 5764 SRTP Extension for DTLS May 2010
As an additional defense against injection into the DTLS handshake channel, the DTLS record type is included in the MAC. Therefore, an SRTP record would be treated as an unknown type and ignored. (See Section 6 of [RFC5246].)
As described in Section 5.1.1, the SRTP and DTLS sequence number spaces are distinct. This means that it is not possible to unambiguously order a given DTLS control record with respect to an SRTP packet. In general, this is relevant in two situations: alerts and rehandshake.
Because DTLS handshake and change_cipher_spec messages share the same sequence number space as alerts, they can be ordered correctly. Because DTLS alerts are inherently unreliable and SHOULD NOT be generated as a response to data packets, reliable sequencing between SRTP packets and DTLS alerts is not an important feature. However, implementations that wish to use DTLS alerts to signal problems with the SRTP encoding SHOULD simply act on alerts as soon as they are received and assume that they refer to the temporally contiguous stream. Such implementations MUST check for alert retransmission and discard retransmitted alerts to avoid overreacting to replay attacks.
Because the rehandshake transition algorithm specified in Section 5.2 requires trying multiple sets of keys if no MKI is used, it slightly weakens the authentication. For instance, if an n-bit MAC is used and k different sets of keys are present, then the MAC is weakened by log_2(k) bits to n - log_2(k). In practice, since the number of keys used will be very small and the MACs in use are typically strong (the default for SRTP is 80 bits), the decrease in security involved here is minimal.
Another concern here is that this algorithm slightly increases the work factor on the receiver because it needs to attempt multiple validations. However, again, the number of potential keys will be very small (and the attacker cannot force it to be larger) and this technique is already used for rollover counter management, so the authors do not consider this to be a serious flaw.
An attacker can impose computational costs on the receiver by sending superficially valid SRTP packets that do not decrypt correctly. In general, encryption algorithms are so fast that this cost is extremely small compared to the bandwidth consumed. The SSRC-DTLS mapping algorithm described in Section 5.1.2 gives the attacker a slight advantage here because he can force the receiver to do more then one decryption per packet. However, this advantage is modest because the number of decryptions that the receiver does is limited by the number of associations he has corresponding to a given destination host/port, which is typically quite small. For comparison, a single 1024-bit RSA private key operation (the typical minimum cost to establish a DTLS-SRTP association) is hundreds of times as expensive as decrypting an SRTP packet.
Implementations can detect this form of attack by keeping track of the number of SRTP packets that are observed with unknown SSRCs and that fail the authentication tag check. If under such attack, implementations SHOULD prioritize decryption and verification of packets that either have known SSRCs or come from source addresses that match those of peers with which it has DTLS-SRTP associations.
This specification defines new tokens to describe the protocol used in SDP media descriptions ("m=" lines and their associated parameters). The new values defined for the proto field are:
o When a RTP/SAVP or RTP/SAVPF [RFC5124] stream is transported over DTLS with the Datagram Congestion Control Protocol (DCCP), then the token SHALL be DCCP/TLS/RTP/SAVP or DCCP/TLS/RTP/SAVPF respectively.
o When a RTP/SAVP or RTP/SAVPF stream is transported over DTLS with UDP, the token SHALL be UDP/TLS/RTP/SAVP or UDP/TLS/RTP/SAVPF respectively.
The "fmt" parameter SHALL be as defined for RTP/SAVP.
See [RFC5763] for how to use offer/answer with DTLS-SRTP.
This document does not specify how to protect RTP data transported over TCP. Potential approaches include carrying the RTP over TLS over TCP (see [SRTP-NOT-MAND]) or using a mechanism similar to that in this document over TCP, either via TLS or DTLS, with DTLS being used for consistency between reliable and unreliable transports. In
McGrew & Rescorla Standards Track [Page 19]
RFC 5764 SRTP Extension for DTLS May 2010
the latter case, it would be necessary to profile DTLS so that fragmentation and retransmissions no longer occurred. In either case, a new document would be required.
This document adds a new extension for DTLS, in accordance with [RFC5246]: enum { use_srtp (14) } ExtensionType;
This extension MUST only be used with DTLS, and not with TLS [RFC4572], which specifies that TLS can be used over TCP but does not address TCP for RTP/SAVP.
Section 4.1.2 requires that all SRTPProtectionProfile values be defined by RFC 5226 "Specification Required". IANA has created a DTLS SRTPProtectionProfile registry initially populated with values from Section 4.1.2 of this document. Future values MUST be allocated via the "Specification Required" profile of [RFC5226].
This specification updates the "Session Description Protocol (SDP) Parameters" registry as defined in Section 8.2.2 of [RFC4566]. Specifically, it adds the following values to the table for the "proto" field.
Type SDP Name Reference ---- ------------------ --------- proto UDP/TLS/RTP/SAVP [RFC5764] proto DCCP/TLS/RTP/SAVP [RFC5764]
proto UDP/TLS/RTP/SAVPF [RFC5764] proto DCCP/TLS/RTP/SAVPF [RFC5764]
IANA has registered the "EXTRACTOR-dtls_srtp" value in the TLS Extractor Label Registry to correspond to this specification.
Special thanks to Flemming Andreasen, Francois Audet, Pasi Eronen, Roni Even, Jason Fischl, Cullen Jennings, Colin Perkins, Dan Wing, and Ben Campbell for input, discussions, and guidance. Pasi Eronen provided Figure 1.
[DTLS1.2] Rescorla, E. and N. Modadugu, "Datagram Transport Layer Security version 1.2", Work in Progress, October 2009.
[RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. Jacobson, "RTP: A Transport Protocol for Real-Time Applications", STD 64, RFC 3550, July 2003.
[RFC4566] Handley, M., Jacobson, V., and C. Perkins, "SDP: Session Description Protocol", RFC 4566, July 2006.
[RFC4572] Lennox, J., "Connection-Oriented Media Transport over the Transport Layer Security (TLS) Protocol in the Session Description Protocol (SDP)", RFC 4572, July 2006.
McGrew & Rescorla Standards Track [Page 21]
RFC 5764 SRTP Extension for DTLS May 2010
[RFC5124] Ott, J. and E. Carrara, "Extended Secure RTP Profile for Real-time Transport Control Protocol (RTCP)- Based Feedback (RTP/SAVPF)", RFC 5124, February 2008.
[RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 5226, May 2008.
[RFC5389] Rosenberg, J., Mahy, R., Matthews, P., and D. Wing, "Session Traversal Utilities for NAT (STUN)", RFC 5389, October 2008.
[RFC5763] Fischl, J., Tschofenig, H., and E. Rescorla, "Framework for Establishing a Secure Real-time Transport Protocol (SRTP) Security Context Using Datagram Transport Layer Security (DTLS)", RFC 5763, May 2010.
[SRTP-NOT-MAND] Perkins, C. and M. Westerlund, "Why RTP Does Not Mandate a Single Security Mechanism", Work in Progress, January 2010.
This section provides a brief overview of Datagram TLS (DTLS) for those who are not familiar with it. DTLS is a channel security protocol based on the well-known Transport Layer Security (TLS) [RFC5246] protocol. Where TLS depends on a reliable transport channel (typically TCP), DTLS has been adapted to support unreliable transports such as UDP. Otherwise, DTLS is nearly identical to TLS and generally supports the same cryptographic mechanisms.
Each DTLS association begins with a handshake exchange (shown below) during which the peers authenticate each other and negotiate algorithms, modes, and other parameters and establish shared keying material, as shown below. In order to support unreliable transport, each side maintains retransmission timers to provide reliable delivery of these messages. Once the handshake is completed, encrypted data may be sent.
Application data is protected by being sent as a series of DTLS "records". These records are independent and can be processed correctly even in the face of loss or reordering. In DTLS-SRTP, this record protocol is replaced with SRTP [RFC3711]
McGrew & Rescorla Standards Track [Page 23]
RFC 5764 SRTP Extension for DTLS May 2010
Appendix B. Performance of Multiple DTLS Handshakes
Standard practice for security protocols such as TLS, DTLS, and SSH, which do inline key management, is to create a separate security association for each underlying network channel (TCP connection, UDP host/port quartet, etc.). This has dual advantages of simplicity and independence of the security contexts for each channel.
Three concerns have been raised about the overhead of this strategy in the context of RTP security. The first concern is the additional performance overhead of doing a separate public key operation for each channel. The conventional procedure here (used in TLS and DTLS) is to establish a master context that can then be used to derive fresh traffic keys for new associations. In TLS/DTLS, this is called "session resumption" and can be transparently negotiated between the peers.
The second concern is network bandwidth overhead for the establishment of subsequent connections and for rehandshake (for rekeying) for existing connections. In particular, there is a concern that the channels will have very narrow capacity requirements allocated entirely to media that will be overflowed by the rehandshake. Measurements of the size of the rehandshake (with resumption) in TLS indicate that it is about 300-400 bytes if a full selection of cipher suites is offered. (The size of a full handshake is approximately 1-2 kilobytes larger because of the certificate and keying material exchange.)
The third concern is the additional round-trips associated with establishing the second, third, ... channels. In TLS/DTLS, these can all be done in parallel, but in order to take advantage of session resumption they should be done after the first channel is established. For two channels, this provides a ladder diagram something like this (parenthetical numbers are media channel numbers)
So, there is an additional 1 RTT (round-trip time) after Channel 1 is ready before Channel 2 is ready. If the peers are potentially willing to forego resumption, they can interlace the handshakes, like so:
In this case, the channels are ready contemporaneously, but if a message in handshake (1) is lost, then handshake (2) requires either a full rehandshake or that Alice be clever and queue the resumption attempt until the first handshake completes. Note that just dropping the packet works as well, since Bob will retransmit.
Authors' Addresses
David McGrew Cisco Systems 510 McCarthy Blvd. Milpitas, CA 95305 USA
EMail: mcgrew@cisco.com
Eric Rescorla RTFM, Inc. 2064 Edgewood Drive Palo Alto, CA 94303 USA