| Network Working Group | M. Sethi |
| Internet-Draft | Ericsson |
| Intended status: Informational | B. Sarikaya |
| Expires: September 10, 2020 | Denpel Informatique |
| D. Garcia-Carrillo | |
| Odin Solutions | |
| March 9, 2020 |
Secure IoT Bootstrapping: A Survey
draft-sarikaya-t2trg-sbootstrapping-08
This draft provides an overview of the various terms that are used when discussing bootstrapping of devices. We document terms that have been used within the IETF as well as other standards bodies. We investigate if the terms refer to the same phenomena or have subtle differences. We provide recommendations on the applicability of terms in different contexts. Finally, this document presents a survey of secure bootstrapping mechanisms available for smart objects that are part of an Internet of Things (IoT) network. The survey does not prescribe any one mechanism and rather presents IoT developers with different options to choose from, depending on their use-case, security requirements, and the user interface available on their smart objects.
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 September 10, 2020.
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://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.
We informally define bootstrapping as "any process that takes place before a device can become operational". While bootstrapping is necessary for all computing devices, until recently, most of our devices typically had sufficient computing power and user interface (UI) for ensuring somewhat smooth operation. For example, a typical laptop device required the user to setup a username/password as well as enter network settings for Internet connectivity. Following these steps ensured that the laptop was fully operational.
The problem of bootstrapping is however exacerbated for Internet of Things (IoT) networks. The size of an IoT network varies from a couple of devices to tens of thousands, depending on the application. Smart objects/things/devices in IoT networks are produced by a variety of vendors and are typically heterogeneous in terms of the constraints on their power supply, communication capability, computation capacity, and user interfaces available. This problem of bootstrapping in IoT was identified by Sethi et al. while developing a bootstrapping solution for smart displays. Although this document focuses on bootstrapping terminology and methods for IoT devices, we do not exclude bootstrapping related terminology used in other contexts.
Bootstrapping devices typically also involves providing them with some sort of network connectivity. Indeed, the functionality of a disconnected device is rather limited. Ipso facto, bootstrapping assumes that some network has been setup a-priori. Setting up a network itself is challenging. For example, the Wi-Fi Alliance Simple Configuration specification [simpleconn] helps users setup home networks. This document is only focused on terminology associated with bootstrapping devices and excludes any discussion on setting up networks.
In addition to our informal definition, it is helpful to look at other definitions of bootstrapping. The IoT@Work project defines bootstrapping in the context of IoT as "the process by which the state of a device, a subsystem, a network, or an application changes from not operational to operational" [iotwork]. Vermillard [vermillard] , on the other hand, describes bootstrapping as the procedure by which an IoT device gets the URLs and secret keys for reaching the necessary servers. Vermillard notes that the same process is useful for re-keying, upgrading the security schemes, and for redirecting the IoT devices to new servers.
There are several terms that have often been used in the context of bootstrapping:
We attempt to find out whether all these terms refer to the same phenomena. We begin by looking at how these terms have been used in various standards and standardization bodies in Section 3. We then summarize our understanding in Section 4, and provide our recommendations on their usage in Section 5. Section 6 provides a taxonomy of bootstrapping methods and Section 7 categorizes methods according to the taxonomy.
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].
Device provisioning protocol (DPP) [dpp] describes itself as a standardized protocol for providing user friendly Wi-Fi setup while maintaining or increasing the security. DPP relies on a configurator, e.g. a smartphone application, for setting up all all other devices, called enrollees, in the network. DPP has the following three phases/sub-protocols:
The OMA LWM2M specification [oma] defines an architecture where a new device (LWM2M client) contacts a Bootstrap-server which is responsible for "provisioning" essential information such as credentials. After receiving this essential information, the LwM2M client device "registers" itself with one or more LwM2M Servers which will manage the device during its lifecycle. The current standard defines the following four bootstrapping modes:
The Open Connectivitiy Foundation [ocf] defines the process before a device is operational as onboarding. The first step of this onboarding process is "configuring" the ownership, i.e., establishing a legitimate user that owns the device. For this, the user is supposed to use an Onboarding tool (OBT) and an Owner Transfer Methods (OTM).
The Onboarding tool (OBT) is described as a logical entity that may be implemented on a single or multiple entities such as a home gateway, a device management tool, etc. OCF lists several optional Owner Transfer Methods (OTMs). At the end of the execution of an OTM, the Onboarding tool must have "provisioned" an Owner Credential onto the device. The following owner transfer methods are specified:
Once the onboarding tool and the new device have authenticated and established secure communication, the onboarding tool "provisions"/"configures" the device with Owner credentials. Owner credentials may consist of certificates, shared keys, or Kerberos tickets for example.
The OBT will ....
The ANIMA working group is working on a bootstrapping solution for devices that relies on 802.1AR vendor certificates called Bootstrapping Remote Secure Key Infrastructures (BRSKI) [I-D.ietf-anima-bootstrapping-keyinfra]. In addition to vendor installed IEEE 802.1AR certificates, a vendor based service on the Internet is required. Before being authenticated, a new device only needs link-local connectivity, and does not require a routable address. When a vendor provides an Internet based service, devices can be forced to join only specific domains. The document highlights that the described solution is aimed in general at non-constrained (i.e. class 2+ defined in [RFC7228]) devices operating in a non-challenged network. It claims to scale to thousands of devices located in hostile environments, such as ISP provided CPE devices which are drop-shipped to the end user.
[RFC8572] defines a bootstrapping strategy for enabling devices to securely obtain all the configuration information with no installer input, beyond the actual physical placement and connection of cables. Their goal is to enable a secure NETCONF [RFC6241] or RESTCONF [RFC8040] connection to the deployment specific network management system (NMS). This bootstrapping method requires the devices to be configured with trust anchors in the form of X.509 certificates. [RFC8572] is similar to BRSKI based on [RFC8366], but using a different set of assumptions ....
There are several stages before a device becomes fully operational. This typically involves establishing some initial trust after which credentials and other parameter.
TBD
While some bootstrapping approaches are more user-intensive and require extensive user-involvement by scanning QR codes or entering passwords; others maybe more automated, such as those that rely on mobile networks [I-D.sethi-gba-constrained]. We classify available bootstrapping solutions into the following major categories:
It is important to note here that categorization of different bootstrapping methods is not always easy or clear. For example, all the opportunistic and leap-of-faith methods become managed methods after the initial vulnerability window. The choice of bootstrapping method used for devices depends heavily on the business case. Questions that may govern the choice include: What third parties are available? Who wants to retain control or avoid work? In each category, there are many different methods of secure bootstrapping available. The choice of the method may also be governed by the type of device being bootstrapped. Depending on the link-layer technology used, and the User Interface (UI) available, one or more of the above mentioned methods might be suitable.
In this section we look at some of the recent bootstrapping proposals for IoT devices both at the IETF and elsewhere. Needless to say, if the devices are capable in terms of their computation power and UI available, they can always rely on many existing methods such as username and password combinations and various EAP methods.
We first discuss some examples of managed bootstrapping methods.
EAP-TLS is a widely used EAP method for network access authentication [RFC7250]. It allows mutual authentication and distributes the keying material for secure subsequent communications. However it only supports certificate-based mutual authentication, and therefore a public key infrastructure is required. The ZigBee Alliance has specified an IPv6 stack aimed at IEEE 802.15.4 [IEEE802.15.4] devices mainly used in smart meters developed primarily for SEP 2.0 (Smart Energy Profile) application layer traffic [SEP2.0]. The ZigBee IP stack uses EAP-TLS for secure bootstrapping of devices.
EAP-PSK [RFC4764] is another EAP method. It realizes mutual authentication and session key derivation using a Pre-Shared Key (PSK). Normally four messages are exchanged in the authentication process. Once the authentication is successful, EAP-PSK provides a protected communication channel. Given the light-weight nature of EAP-PSK, it can often be a good choice on constrained devices.
CoAP-EAP [I-D.marin-ace-wg-coap-eap] defines a bootstrapping service for IoT. They propose the transport of EAP over CoAP [RFC7252] for the constrained link, and communication with AAA infrastructures in the non-constrained link to provide scalability among other characteristics. Upon a successful authentication, key material is derived to protect CoAP messages exchanged between the smart object and the authenticator. They discuss the use of EAP-PSK in the draft, but state that, since they are specifying a new EAP lower layer, any EAP method that results in generation of cryptographic material is suitable.
Protocol for Carrying Authentication for Network Access (PANA) [RFC5191] is a network layer protocol with which a node can authenticate itself to gain access to the network. PANA does not define a new authentication protocol and rather uses EAP over User Datagram Protocol (UDP) for authentication. Colin O'Flynn [I-D.oflynn-core-bootstrapping] proposes the use of PANA for secure bootstrapping of resource constrained devices. He demonstrates how a 6LowPAN Border Router (PANA Authentication Agent (PAA)) can authenticate the identity of a joining constrained device (PANA Client). Once the constrained device has been successfully authenticated, the border router can also provide network and security parameters to the joining device. Hernandez-Ramos et al. [panaiot] also use EAP-TLS over PANA for secure bootstrapping of smart objects. They also extend their bootstrapping scheme for configuring additional keys that are used for secure group communication.
When a device is not a direct neighbor of the authenticator, its parent node MUST act as relay. Different EAP encapsulation protocols have different mechanisms for the relay function, such as the PANA Relay Element (PRE).
After a successful bootstrapping, the device runs neighbor discovery protocol to get an IPv6 address assigned [RFC6775]. Data transfer can be secured using DTLS or IPSec. Keys derived from EAP TLS are used in either generating DTLS ciphering keys after a successful DTLS handshake or IPSec ESP ciphering keys after a successful IKEv2 handshake.
Generic Bootstrapping Architecture (GBA) is another bootstrapping method that falls in centralized category. GBA is part of the 3GPP standard [TS33220] and is based on 3GPP Authentication and Key Agreement (3GPP AKA). GBA is an application independent mechanism to provide a client application (running on the User equipment (UE)) and any application server with a shared session secret. This shared session secret can subsequently be used to authenticate and protect the communication between the client application and the application server. GBA authentication is based on the permanent secret shared between the UE's Universal Integrated Circuit Card (UICC), for example SIM card, and the corresponding profile information stored within the cellular network operator's Home Subscriber System (HSS) database. [I-D.sethi-gba-constrained] describes a resource-constrained adaptation of GBA to IoT applications.
The four bootstrapping modes specified by the Open Mobile Alliance (OMA) Light-weight M2M (LwM2M) standard require some sort of pre-provisioned credentials on the device. All the four modes are examples of managed bootstrapping methods.
The Kerberos protocol [RFC4120] is a network authentication protocol that allows several endpoints to communicate over an insecure network. Kerberos relies on a symmetric cryptography scheme and requires a trusted third party, that guarantees the identities of the various actors. It relies on the use of "tickets" for nodes to prove identity to one another in a secure manner. There has been research work on using Kerberos for IoT devices [kerberosiot].
It is also important to mention some of the work done on implicit certificates and identity-based cryptographic schemes [himmo], [implicit]. While these are interesting and novel schemes that can be a part of securely bootstrapping devices, at this point, it is hard to speculate on whether such schemes would see large-scale deployment in the future.
Low Power Wide Area Network (LPWAN) encompasses a wide variety of technologies, generally, with severe constraints in the link in comparison with other typical IoT technologies such as Bluetooth or IEEE 802.15.4. LPWAN typically presents a star topology with support for thousands of devices per antenna.
Among the wide variety of technologies considered as part of LPWAN, we highlight the ones mentioned in the LPWAN overview document of the LPWAN working group [RFC8376]: LoRaWAN, Narrowband IoT (NB-IoT), SIGFOX and Wi-SUN Alliance Field Area Network (FAN). Each technology has different methods to provide security for the communications. Bootstrapping is not directly tackled by all of them, having in some cases proprietary solutions that are not publicly accessible and in other cases key distribution is not even considered. Among the previous LPWAN technologies, bootstrapping is considered in Wi-SUN Alliance Field Area Network (FAN) and LoRaWAN provides Joining process to derive key material based on some previous key material installed in the device.
Following the definition in Section 6 we find that they all fall into the managed classification. This is because in one way or another, a previous trust relationship has been established and authentication credentials have been installed in the devices.
In short, LPWANs are still under development, and as it is identified in [RFC8376], due to the characteristics of these technologies, they are prime candidates to benefit from a standardized Authentication, Accounting, and Authorization (AAA) infrastructure [RFC2904] as a way of offering a scalable solution for some of the security and management issues that are present in LPWANs.
While managed methods are viable for many IoT devices, they may not be suitable or desirable in all scenarios. All the managed methods assume that some credentials are provisioned into the device. These credentials may be in the device micro-controller or in a replaceable smart card such as a SIM card. The methods also sometimes assume that the manufacturer embeds these credentials during the device manufacture on the factory floor. However, in many cases the manufacturer may not have sufficient incentive to do this. In other scenarios, it may be hard to completely trust and rely on the device manufacturer to securely perform this task. Therefore, many times, P2P or Adhoc methods of bootstrapping are used. We discuss a few example next.
P2P or ad-hoc bootstrapping methods are used for establishing keys and credential information for secure communication without any pre-provisioned information. These bootstrapping mechanisms typically rely on an out-of-band (OOB) channel in order to prevent man-in-the-middle (MitM) attacks. P2P and ad-hoc methods have typically been used for securely pairing personal computing devices such as smart phones. [devicepairing] provides a survey of such secure device pairing methods. Many original pairing schemes required the user to enter the same key string or authentication code to both devices or to compare and approve codes displayed by the devices. While these methods can provide reasonable security, they require user interaction that is relatively unnatural and often considered a nuisance. Thus, there is ongoing research for more natural ways of pairing devices. To reduce the amount of user-interaction required in the pairing process, several proposals use contextual or location-dependent information, or natural user input such as sound or movement, for device pairing [proximate].
The local association created between two devices may later be used for connecting/introducing one of the devices to a centralized server. Such methods would however be classified as hybrids.
EAP-NOOB [I-D.aura-eap-noob] is an example of P2P and ad-hoc bootstrapping method that establishes a security association between an IoT device (node) and an online server (unlike pairing two devices for local connections over WiFi or Bluetooth).
EAP-NOOB defines an EAP method where the authentication is based on a user-assisted out-of-band (OOB) channel between the server and peer. It is intended as a generic bootstrapping solution for Internet-of-Things devices which have no pre-configured authentication credentials and which are not yet registered on the authentication server. This method claims to be more generic than most ad-hoc bootstrapping solutions in that it supports many types of OOB channels. The exact in-band messages and OOB message contents are specified and not the OOB channel details. Also, EAP-NOOB supports IoT devices with only output (e.g. display) or only input (e.g. camera). It makes combined use of both secrecy and integrity of the OOB channel for more robust security than the ad-hoc solutions.
Thread Group commissioning [threadcommissioning] introduces a two phased process i.e. Petitioning and Joining. Entities involved are Leader, Joiner, Commissioner, Joiner Router and Border Router. Leader is the first device in Thread Network that must be commissioned using out-of-band process and is used to inject correct user generated Commissioning Credentials (can be changed later) into Thread Network. Joiner is the node that intends to get authenticated and authorized on Thread Network. Commissioner is either within the Thread Network (Native) or connected with Thread Network via a WLAN (External).
Under some topologies, Joiner Router and Border Router facilitate the Joiner node to reach Native and External Commissioner, respectively. Petitioning begins before Joining process and is used to grant sole commissioning authority to a Commissioner. After an authorized Commissioner is designated, eligible thread devices can join network. Pair-wise key is shared between Commissioner and Joiner, network parameters (Network Name, Security Policy, Steering Data, Commissioning Data Timestamp, Commissioning Credential, Network Master Key, Network Key Sequence, Network Mesh-Local ULA, Border Router Locator, Commissioner Session ID, XPANID, PANID and Channel) are sent out securely (using pair-wise key) by Joiner Router to Joiner for letting Joiner to join the Thread Network. Entities involved in Joining process depends on system topology i.e. location of Commissioner and Joiner.
Thread networks only operates using IPv6. Thread devices can devise GUAs (Global Unicast Addresses) [RFC4291]. Provision also exist via Border Router, for Thread device to acquire individual global address by means of DHCPv6 or using SLAAC (Stateless Address Autoconfiguration) address derived with advertised network prefix.
DPP bootstrapping explained earlier is an example of ad-hoc bootstrapping method.
Next, we look at a leap-of-faith/opportunistic bootstrapping method for IoT devices.
Bergmann et al. [simplekey] develop a secure bootstrapping mechanism that does not rely on pre-provisioned credentials using resurrecting-duckling imprinting scheme. Their bootstrapping protocol involves three distinct phases: discover (the duckling node searches for network nodes that can act as mother node), imprint (the mother node imprints a shared secret establishing a secure channel once a positive response is received for the imprinting request) and configure (additional configuration information such as network prefix and default gateway are configured). In this model for bootstrapping, a small initial vulnerability window is acceptable and can be mitigated using techniques such as a Faraday Cage (securing the communication physically) to protect the environment of the mother and duck nodes, though this may be inconvenient for the user.
[RFC7250] defines how raw public keys can be used to authenticate constrained devices for mutual authentication using EAP-TLS or DTLS. Raw public key TLS/DTLS extension simplifies client_certificate_type and server_certificate_type to carry only SubjectPublicKeyInfo structure with the raw public key instead of many other parameters found in the certificates. The device and the authentication server (AS) exchange client_hello and server_hello messages and send their raw public keys. The device and AS validate the keys by comparing the pre-configured values [I-D.sarikaya-6lo-bootstrapping-solution]. This bootstrapping method can be seen as a hybrid. This is because it generally requires an out-of-band (OOB) step (P2P/Ad-hoc) where the raw public keys [RFC7250] are provided to the authenticating entities, after which the actual authentication occurs online (managed). Raw public key approach when used with DTLS offers a simple secure bootstrapping solution especially for smart energy and building automation applications. It can be easily integrated with the Constrained Application Protocol (CoAP).
Bootstrapping protocols that do not rely on a pre-shared key for peer authentication generally rely on an online or offline third-party (e.g., an authentication server, a key distribution center in Kerberos, a certification authority in PKI, a private key generator in ID-based cryptography and so on) to prevent man-in-the-middle attacks during bootstrapping. Depending on use cases, a resource-constrained device may not always have access to an online third-party for bootstrapping. Some bootstrapping methods therefore rely on a configuration tool (such as a smartphone) that assists the bootstrapping process by providing temporary reachability to the online server.
Depending on use cases, a bootstrapping protocol may deal with authorization separately from authentication in terms of timing and signaling path. For example, two resource-constrained devices A and B may perform mutual authentication using authentication credentials provided by an offline third-party X whereas resource-constrained device A obtains authorization for running a particular application with resource-constrained device B from an online third-party Y before or after the authentication. In some use cases, authentication and authorization are tightly coupled, e.g., successful authentication also means successful authorization.
If authorization information communicated includes cryptographic keys, care must be taken for provisioning the keys, e.g., guidelines for AAA-based key management are described in [RFC4962]. Re-bootstrapping of IoT devices may required and therefore there must be adequate provisions for revocation and re-bootstrapping of authentication/authorization credentials. Re-bootstrapping must be as secure as the initial bootstrapping regardless of whether this re-bootstrapping is done manually or automatically over the network.
If resource-constrained devices use a multicast group key for authentications of peers that belong to the group, or for message authentication/encryption, the group key must be securely distributed to the current members of the group. Protocols designed for group key management [RFC4046] may be used for group key distribution after the initial bootstrapping. Alternatively, key wrap attributes for securely encapsulating group key may be defined in network access authentication protocols such as PANA. Those protocols use an end-to-end, point-to-point communication channel with a pair-wise security association between a key distribution center and each key recipient. Further considerations may be needed for more efficient group key management to support a large number of resource-constrained devices.
There are no IANA considerations for this document.
We would like to thank Tuomas Aura, Hannes Tschofenig, and Michael Richardson for providing extensive feedback as well as Rafa Marin for his support.