| DICE Working Group | S. Kumar, Ed. |
| Internet-Draft | Philips Research |
| Intended status: Standards Track | S. Keoh |
| Expires: April 23, 2015 | University of Glasgow Singapore |
| O. Garcia-Morchon | |
| Philips Research | |
| October 20, 2014 |
DTLS Relay for Constrained Environments
draft-kumar-dice-dtls-relay-02
The 6LoWPAN and CoAP standards defined for resource-constrained devices are fast emerging as the de-facto protocols for enabling the Internet-of-Things (IoTs). Security is an important concern in IoTs and the DTLS protocol has been chosen as the preferred method for securing CoAP messages. DTLS is a point-to-point protocol relying on IP routing to deliver messages between the client and the server. However in some low-power lossy networks (LLNs) with multi-hop, a new "joining" device may not be initially IP-routable. Moreover, it exists in a separate, unauthenticated domain at the point of first contact and therefore cannot be initially trusted. This puts limitations on the ability to use DTLS as an authentication and confidentiality protocol at this stage. These devices being Resource-constrained often cannot accommodate more than one security protocol in their code memory. To overcome this problem we suggest DTLS as the single protocol and therefore, we present a DTLS Relay solution for the non-IP routable "joining" device to enable it to establish a secure DTLS connection with a DTLS Server. Furthermore we present a stateful and stateless mode of operation for the DTLS Relay.
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 http://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 April 23, 2015.
Copyright (c) 2014 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.
For the Internet of Things (IoT) to become a reality, it will require the participation of constrained nodes in constrained networks [RFC7228]. These constrained nodes typically implement the IPv6 over Low-Power Wireless Personal Area Networks (6LoWPAN) [RFC4944] and Constrained Application Protocol (CoAP) [RFC7252] standards. The 6LoWPAN adaptation layer allows for transmission of IPv6 Packets over IEEE 802.15.4 networks [ieee802.15.4], thereby enabling end-to-end IPv6 connectivity between constrained nodes and other devices on the Internet. CoAP is a web protocol based on REST architecture designed for constrained node networks. It supports binding to UDP [RFC0768], which has advantages over TCP [RFC0793] when used in low-power lossy networks (LLNs) such as IEEE 802.15.4 [ieee802.15.4].
Security is an important concern in such a constrained node network, which could be used in various application domains such as smart energy and building automation. However, security protocols are often heavy-weight in terms of both code and network processing. Use of multiple security protocols for different purposes and at different networking layers is problematic in constrained devices, therefore the use of a single security protocol to fulfil multiple security requirements is greatly preferred.
CoAP has chosen Datagram Transport Layer Security (DTLS) [RFC6347] as the preferred security protocol for authenticity and confidentiality of the messages. It is based on Transport Layer Security (TLS) [RFC5246] with modifications to run over UDP. DTLS makes use of additional reliability mechanisms in its handshake due to the lack of TCP reliable transmission mechanisms that are available to TLS.
DTLS is a client-server protocol relying on the underlying IP layer to perform the routing between the DTLS Client and the DTLS Server. However in some LLNs with multi-hop, a new "joining" device may not be initially IP routable until it is authenticated to the network. A new "joining" device can only initially use a link-local IPv6 address to communicate with a neighbour node using neighbour discovery [RFC6775] until it receives the necessary network configuration parameters. However, before the device can receive these configuration parameters, it may need to authenticate itself or wish to authenticate the network to which it connects. Although DTLS is a suitable protocol for such authentication and secure transfer of configuration parameters, it would not work due to the lack of IP routability of DTLS messages between DTLS Client and DTLS Server.
We present a DTLS Relay solution to overcome this problem for the "joining" device to establish a DTLS connection with a DTLS Server. This draft is inspired by the Protocol for carrying Authentication for Network Access (PANA) Relay Element [RFC6345] which is intended to solve a similar problem when PANA [RFC5191] is used as the transport protocol for Extensible Authentication Protocol (EAP) [RFC3748] based network access. Recently there has been interest in transporting EAP over CoAP [I-D.marin-ace-wg-coap-eap][I-D.ohba-core-eap-based-bootstrapping] and presented DTLS Relay solution can be used to secure these messages. Further, we present a stateful and stateless mode of operation for the DTLS Relay.
This draft is an early description of the solutions and does not provide the complete details yet. This draft is structured as follows: we present a use-case for the DTLS Relay in Section 2, then present the DTLS Relay solution in Section 3 for stateful and stateless mode of operation. We compare these two solutions in Section 3.3. Further we present some security considerations in Section 5.
We present here a target usecase based on [I-D.jennings-core-transitive-trust-enrollment] describing a rendezvous protocol that allows a constrained IoT device to securely connect into a system or network. The main idea is that the joining Device has a pre-established trust relationship with a "Transfer Agent" entity, for e.g. Pre-Shared Keys provisioned during manufacturing. This "Transfer Agent" provides the needed trust credentials to the Device and/or a Controller in the system to establish a secured connection to perform further authentication and transfer of system/network configuration parameters. This step is enabled by an "Introducer" entity which informs the "Transfer Agent" about the details of Controller to which the joining Device should connect, and provide to the Controller the identity including one-time credentials for enable secure connection to the Device. The transitive trust trust establishment procedure is explained in detail in [I-D.jennings-core-transitive-trust-enrollment] and we focus here on how to enable this using DTLS.
As depicted in the Figure 1, the joining Device (D) is more than one hop away from the Controller (C) and not yet authenticated into the network. At this stage, it can only communicate one-hop to its nearest neighbour (N) using their link-local IPv6 addresses. However, the Device needs to communicate with end-to-end security with a Transfer Agent (T) or a Controller (C) to authenticate and get the relevant system/network parameters. If the Device (D) initiates a DTLS connection to the Transfer Agent whose IP address has been pre-configured, then the packets are dropped at the neighbour (N) since the Device (D) is not yet admitted to the network or there is no IP routability to Device (D) for any returned messages.
Transfer Agent
++++
|T |
| | +--+
++++ |N'|
| --+--+
| ++++ /
| |C |---- +--+ +--+
--| | \ |N |........|D |
++++ \-----| | | |
Controller +--+ +--+
Neighbour "Joining" Device
Figure 1: Use case depiction in a multi-hop network
Furthermore, the Device (D) may wish to establish a secure connection to the Controller (C) in the network assuming appropriate credentials are exchanged out-of-band, e.g. a hash of the Device (D)'s raw public key could be provided to the Controller (C). However, the Device (D) is unaware of the IP address of the Controller (C) to initiate a DTLS connection and perform authentication with.
To overcome these problems with non-routability of DTLS packets and/or discovery of the destination address of the DTLS Server to contact, we define a DTLS Relay solution. This DTLS Relay ability is configured into all authenticated devices in the network which may act as the Neighbour (N) device for newly joining nodes. The DTLS Relay allows for relaying of the packets from the Neighbour (N) using IP routing to the intended DTLS Server. Furthermore, if the DTLS Server address is not known to the joining Device (D), then messages are delivered to a pre-configured DTLS Server address (most likely the Controller (C)) known to the DTLS Relay.
In this section, we describe how the DTLS Relay functionality can be achieved. When a joining device as a client attempts a DTLS connection (for example to a "Transfer Agent"), it uses its link-local IP address as its IP source address. This message is transmitted one-hop to a neighbour node. Under normal circumstances, this message would be dropped at the neighbour node since the joining device is not yet IP routable or it is not yet authenticated to send messages through the network. However, if the neighbour device has the DTLS Relay functionality enabled, it relays the DTLS message to a specific DTLS Server. Additional security mechanisms need to exist to prevent this relaying functionality being used by rogue nodes to bypass any network authentication procedures. These mechanisms are discussed in Section 5.
The DTLS Relay can operate in two different modes: stateful and stateless. We present here both modes, however for inter-operability, only one of the modes should be mandated. Within each mode, the DTLS Relay can further relay packets based on the client-defined DTLS Server address or a DTLS Server address that has been configured into the DTLS Relay.
On receiving a DTLS message from a joining device, the neighbour node enters into DTLS Relay stateful mode. In this mode, the neighbour node has the additional DTLS Relay functionality to send DTLS messages further to the end-point DTLS Server the joining device wishes to contact. In the stateful mode of operation, the message is transmitted to the end-point DTLS Server as if it originated from the DTLS Relay, by replacing the IP address and port to the DTLS Relay's own IP address and a randomly chosen port. The DTLS message itself is not modified.
Additionally, the DTLS Relay must track the ongoing DTLS connections based on the following 4-tuple stored locally:
The DTLS Server communicates with the DTLS Relay as if it were communicating with the DTLS Client, without any modification required to the DTLS messages. On receiving a DTLS message from the DTLS Server, the DTLS Relay looks up its locally stored 4-tuple array to identify to which DTLS Client (if multiple exist) the message belongs. The DTLS message's destination address and port are replaced with the link-local address and port of the corresponding DTLS Client respectively and the DTLS message is then forwarded to the DTLS Client. The DTLS Relay does not modify the DTLS packets and therefore the normal processing and security of DTLS is unaffected.
The following message flow diagram indicates the various steps of the process where the DTLS Server address in known to the joining device:
+---------------+---------------+----------------+---------------------------+ | DTLS Client | DTLS Relay | DTLS Server | Message | | (C) | (R) | (S) | Src_IP:port | Dst_IP:port | +---------------+---------------+----------------+-------------+-------------+ | --ClientHello--> | IP_C:p_C | IP_S:5684 | | --ClientHello--> | IP_R:p_R | IP_S:5684 | | | | | | <--ServerHello-- | IP_S:5684 | IP_R:p_R | | : | | | | <--ServerHello-- | IP_S:5684 | IP_C:p_C | | : | | | | :: | : | : | | :: | : | : | | --Finished--> | IP_C:p_C | IP_S:5684 | | --Finished--> | IP_R:p_R | IP_S:5684 | | | | | | <--Finished-- | IP_S:5684 | IP_R:p_R | | <--Finished-- | IP_S:5684 | IP_C:p_C | | :: | : | : | +------------------------------------------------+-------------+-------------+ IP_C:p_C = Link-local IP address and port of DTLS Client IP_S:5684 = IP address and coaps port of DTLS Server IP_R:p_R = IP address and port of DTLS Relay
Figure 2: Message flow in Stateful mode with DTLS Server defined by DTLS Client
In the situation where the joining device is unaware of the IP address of the DTLS Server it needs to contact, for e.g. the Controller of the network, the DTLS Relay can be configured with IP destination of the default DTLS Server that a DTLS client (joining device) needs to contact. The DTLS client initiates its DTLS request as if the DTLS Relay is the intended end-point DTLS Server. The DTLS Relay changes the IP packet (without modifying the DTLS message) as in the previous case by modifying both the source and destination IP addresses to forward the message to the intended DTLS Server. The DTLS Relay keeps a similar 4-tuple array to enable translation of the DTLS messages received from the DTLS Server and forward it to the DTLS Client. The following message flow indicates this process:
+---------------+---------------+----------------+---------------------------+ | DTLS Client | DTLS Relay | DTLS Server | Message | | (C) | (R) | (S) | Src_IP:port | Dst_IP:port | +---------------+---------------+----------------+-------------+-------------+ | --ClientHello--> | IP_C:p_C | IP_Ra:5684 | | --ClientHello--> | IP_Rb:p_Rb| IP_S:5684 | | | | | | <--ServerHello-- | IP_S:5684 | IP_Rb:p_Rb | | : | | | | <--ServerHello-- | IP_Ra:5684| IP_C:p_C | | : | | | | :: | : | : | | :: | : | : | | --Finished--> | IP_C:p_C | IP_Ra:5684 | | --Finished--> | IP_Rb:p_Rb| IP_S:5684 | | | | | | <--Finished-- | IP_S:5684 | IP_Rb:p_Rb | | <--Finished-- | IP_Ra:5684| IP_C:p_C | | :: | : | : | +------------------------------------------------+-------------+-------------+ IP_C:p_C = Link-local IP address and port of DTLS Client IP_S:5684 = IP address and coaps port of DTLS Server IP_Ra:5684 = Link-local IP address and coaps port of DTLS Relay IP_Rb:p_Rb = IP address (can be same as IP_Ra) and port of DTLS Relay
Figure 3: Message flow in Stateful mode with DTLS Server defined by DTLS Relay
In the alternative mode of operation for the DTLS Relay, a stateless approach is applied where the DTLS Relay does not need to store a local 4-tuple array. Just as in the previous case, if an untrusted DTLS Client that can only use link-local addressing wants to contact a trusted end-point DTLS Server, it send the DTLS message to the DTLS Relay. Instead of changing the IP addresses and port of the IP packet, the DTLS Relay encapsulates this message into a new type of message called DTLS Relay (DRY) message. The DRY message consists of two parts:
On receiving the DRY message, the DTLS Server decapsulates it to retrieve the two parts. It uses the Header field information to transiently store the DTLS Client's address and port. The DTLS Server then performs the normal DTLS operations on the DTLS message from the Contents field. However, when the DTLS Server replies, it also encapsulates its DTLS message in a DRY message back to the DTLS Relay. The Header contains the original source link-local address and port of the DTLS Client from the transient state stored earlier (which can now be discarded) and the Contents field contains the DTLS message.
On receiving the DRY message, the DTLS Relay decapsulates it to retrieve the two parts. It uses the Header field to relay the DTLS message retrieved from the Contents field to the right DTLS Client.
The following figure depicts the message flow diagram when the DTLS Server end-point address is known only to the Relay:
+----------------+---------------------+---------------------+---------------------------+ | DTLS Client | DTLS Relay | DTLS Server | Message | | (C) | (R) | (S) | Src_IP:port | Dst_IP:port | +----------------+---------------------+---------------------+-------------+-------------+ | --ClientHello--> | IP_C:p_C | IP_Ra:5684 | | --DRY[H(IP_C:p_C),C(ClientHello)]--> | IP_Rb:p_Rb| IP_S:5684 | | | | | | <--DRY[H(IP_C:p_C),C(ServerHello)]-- | IP_S:5684 | IP_Rb:p_Rb | | : | | | | <--ServerHello-- | IP_Ra:5684| IP_C:p_C | | : | | | | :: | : | : | | :: | : | : | | --Finished--> | IP_C:p_C | IP_Ra:5684 | | --DRY[H(IP_C:p_C),C(Finished)]--> | IP_Rb:p_Rb| IP_S:5684 | | | | | | <--DRY[H(IP_C:p_C),C(Finished)]-- | IP_S:5684 | IP_Rb:p_Rb | | <--Finished-- | IP_Ra:5684| IP_C:p_C | | :: | : | : | +------------------------------------------------------------+-------------+-------------+ IP_C:p_C = Link-local IP address and port of DTLS Client IP_S:5684 = IP address and coaps port of DTLS Server IP_Ra:5684 = Link-local IP address and coaps port of DTLS Relay IP_Rb:p_Rb = IP address(can be same as IP_Ra) and port of DTLS Relay DRY[H(),C()] = DTLS Relay message with header H and content C
Figure 4: Message flow in Stateless mode with DTLS Server defined by DTLS Relay
The message flow for the case in which the DTLS Client is aware of the end-point DTLS Server's IP address is similar and not described further. It can be derived based on Figure 2 and Figure 4.
The stateful and stateless mode of operation for the DTLS Relay have their advantages and disadvantages. This comparison should enable to make a good choice between the two based on the available device resources and network bandwidth in a given deployment.
+-------------------+-----------------------------------+--------------------------------+ | Properties | Stateful mode | Stateless mode | +-------------------+-----------------------------------+--------------------------------+ | State information |The Relay needs additional storage | No information is maintained by| | |to maintain mapping of the joining | the Relay. | | |device’s address with the port | | | |number being used to communicate | | | |with the Server. | | +-------------------+-----------------------------------+--------------------------------+ | Packet size |The size of the relayed message is |The size of the relayed message| | |the same as the original message . |is bigger than the original, it | | | |includes additional source and | | | |destination addresses. | +-------------------+-----------------------------------+--------------------------------+ | Standardization |The additional functionalities of |New DRY message to encapsulate | | requirements |the Relay to maintain state |DTLS message. The Server and the| | |information, and modify the source |Relay have to understand the DRY| | |and destination addresses of the |message in order to process it. | | |DTLS handshake messages. | | +-------------------+-----------------------------------+--------------------------------+ Table 1: Comparison between stateful and stateless mode DTLS Relay
Figure 5
tbd
Note to RFC Editor: this section may be removed on publication as an RFC.
Additional security considerations need to be taken into account when forwarding messages from devices through a network to which it has not yet been admitted since this can lead to denial-of-service (DoS) attacks or misuse of network resources without proper authentication. There are various solution options by which one could try to limit the damage that an attacker can cause by DoS. One way to overcome any large scale misuse of the network is to have a management message from the Controller that initiates already authenticated devices in the network to enable or disable the DTLS Relay mode. This is often possible since the administrator of the network is aware when new devices join the network either because of the "Introduction" phase or commissioning phase. Alternatively the management message can be used to control a different networking layer on the relay nodes that disable new nodes from joining. Such solution options are orthogonal to the DTLS relay functionality and should be built in based on the underlying network capabilities and deployment scenario.
In terms of the two different modes, Stateful mode has additional security issues since the DTLS Relay has to store state from an unauthenticated node and then relay a message, expecting a corresponding reply sometime in the future. Furthermore, the DTLS Server has to store state as well but it is more transient. This could lead to a simple localised attack on a DTLS Relay whereby a rogue device could use up state storage on a DTLS Relay quite easily, thus denying a legitimate device from being able to gain access.
In comparison, in the Stateless mode the DTLS Relay does not store any state, and therefore an attack as described above is not possible. Also a DTLS Server can legitimately silently discard a DTLS message without concern as the DTLS Relay has no further knowledge or state stored of the DTLS Client. The DTLS cookie mechanism is a good addition to a stateless transaction which improves the likelihood a DTLS Server is talking to a genuine DTLS Client.
The authors would like to thank Sahil Sharma, Ernest Ma, Dee Denteneer, Peter Lenoir and Martin Turon for the valuable discussions. Also thank Robert Craige for his valuable comments and edits.
| [RFC6347] | Rescorla, E. and N. Modadugu, "Datagram Transport Layer Security Version 1.2", RFC 6347, January 2012. |