| Routing Over Low-Power and Lossy Networks | T. Tsao |
| Internet-Draft | R. Alexander |
| Intended status: Informational | Cooper Power Systems |
| Expires: March 12, 2015 | M. Dohler |
| CTTC | |
| V. Daza | |
| A. Lozano | |
| Universitat Pompeu Fabra | |
| M. Richardson, Ed. | |
| Sandelman Software Works | |
| September 8, 2014 |
A Security Threat Analysis for Routing Protocol for Low-power and lossy networks (RPL)
draft-ietf-roll-security-threats-10
This document presents a security threat analysis for the Routing Protocol for Low-power and lossy networks (RPL, ROLL). The development builds upon previous work on routing security and adapts the assessments to the issues and constraints specific to low-power and lossy networks. A systematic approach is used in defining and evaluating the security threats. Applicable countermeasures are application specific and are addressed in relevant applicability statements.
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 RFC 2119 [RFC2119].
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 March 12, 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.
In recent times, networked electronic devices have found an increasing number of applications in various fields. Yet, for reasons ranging from operational application to economics, these wired and wireless devices are often supplied with minimum physical resources; the constraints include those on computational resources (RAM, clock speed, storage), communication resources (duty cycle, packet size, etc.), but also form factors that may rule out user access interfaces (e.g., the housing of a small stick-on switch), or simply safety considerations (e.g., with gas meters). As a consequence, the resulting networks are more prone to loss of traffic and other vulnerabilities. The proliferation of these low-power and lossy networks (LLNs), however, are drawing efforts to examine and address their potential networking challenges. Securing the establishment and maintenance of network connectivity among these deployed devices becomes one of these key challenges.
This document presents a threat analysis for securing the Routing Protocol for LLNs (RPL). The process requires two steps. First, the analysis will be used to identify pertinent security issues. The second step is to identify necessary countermeasures to secure RPL. As there are multiple ways to solve the problem and the specific tradeoffs are deployment specific, the specific countermeasure to be used is detailed in applicability statements.
This document uses [IS07498-2] model, which describes Authentication, Access Control, Data Confidentiality, Data Integrity, and Non-Repudiation security services and to which Availability is added.
All of this document concerns itself with securing the control plane traffic. As such it does not address authorization or authentication of application traffic. RPL uses multicast as part of it's protocol, and therefore mechanisms which RPL uses to secure this traffic MAY be applicable to MPL control traffic as well: the important part is that the threats are similiar.
ROLL has specified a set of routing protocols for Lossy and Low-resource Networks (LLN) [RFC6550]. A number of applicability texts describes a subset of these protocols and the conditions which make the subset the correct choice. The text recommends and motivates the accompanying parameter value ranges. Multiple applicability domains are recognized including: Building and Home, and Advanced Metering Infrastructure. The applicability domains distinguish themselves in the way they are operated, their performance requirements, and the most probable network structures. Each applicability statement identifies the distinguishing properties according to a common set of subjects described in as many sections.
The common set of security threats herein are referred to by the applicability statements, and that series of documents describes the preferred security settings and solutions within the applicability statement conditions. This applicability statements may recommend more light weight security solutions and specify the conditions under which these solutions are appropriate.
This document adopts the terminology defined in [RFC6550], in [RFC4949], and in [RFC7102].
The terms control plane and forwarding plane are used consistently with section 1 of [RFC6192].
Routing security, in essence, ensures that the routing protocol operates correctly. It entails implementing measures to ensure controlled state changes on devices and network elements, both based on external inputs (received via communications) or internal inputs (physical security of device itself and parameters maintained by the device, including, e.g., clock). State changes would thereby involve not only authorization of injector's actions, authentication of injectors, and potentially confidentiality of routing data, but also proper order of state changes through timeliness, since seriously delayed state changes, such as commands or updates of routing tables, may negatively impact system operation. A security assessment can therefore begin with a focus on the assets [RFC4949] that may be the target of the state changes and the access points in terms of interfaces and protocol exchanges through which such changes may occur. In the case of routing security, the focus is directed towards the elements associated with the establishment and maintenance of network connectivity.
This section sets the stage for the development of the analysis by applying the systematic approach proposed in [Myagmar2005] to the routing security, while also drawing references from other reviews and assessments found in the literature, particularly, [RFC4593] and [Karlof2003]. The subsequent subsections begin with a focus on the elements of a generic routing process that is used to establish routing assets and points of access to the routing functionality. Next, the [ISO.7498-2.1988] security model is briefly described. Then, consideration is given to issues specific to or amplified in LLNs. This section concludes with the formulation of a set of security objectives for RPL.
An asset is an important system resource (including information, process, or physical resource), the access to, corruption or loss of which adversely affects the system. In the control plane context, an asset is information about the network, processes used to manage and manipulate this data, and the physical devices on which this data is stored and manipulated. The corruption or loss of these assets may adversely impact the control plane of the network. Within the same context, a point of access is an interface or protocol that facilitates interaction between control plane assets. Identifying these assets and points of access will provide a basis for enumerating the attack surface of the control plane.
A level-0 data flow diagram [Yourdon1979] is used here to identify the assets and points of access within a generic routing process. The use of a data flow diagram allows for a clear and concise model of the way in which routing nodes interact and process information, and hence provides a context for threats and attacks. The goal of the model is to be as detailed as possible so that corresponding assets, points of access, and process in an individual routing protocol can be readily identified.
Figure 1 shows that nodes participating in the routing process transmit messages to discover neighbors and to exchange routing information; routes are then generated and stored, which may be maintained in the form of the protocol forwarding table. The nodes use the derived routes for making forwarding decisions.
...................................................
: :
: :
|Node_i|<------->(Routing Neighbor _________________ :
: Discovery)------------>Neighbor Topology :
: -------+--------- :
: | :
|Node_j|<------->(Route/Topology +--------+ :
: Exchange) | :
: | V ______ :
: +---->(Route Generation)--->Routes :
: ---+-- :
: | :
: Routing on a Node Node_k | :
...................................................
|
|Forwarding |
|On Node_l|<-------------------------------------------+
Notation:
(Proc) A process Proc
________
topology A structure storing neighbor adjacency (parent/child)
--------
________
routes A structure storing the forwarding information base (FIB)
--------
|Node_n| An external entity Node_n
-------> Data flow
Figure 1: Data Flow Diagram of a Generic Routing Process
It is seen from Figure 1 that
A focus on the above list of assets and points of access enables a more directed assessment of routing security; for example, it is readily understood that some routing attacks are in the form of attempts to misrepresent routing topology. Indeed, the intention of the security threat analysis is to be comprehensive. Hence, some of the discussion which follows is associated with assets and points of access that are not directly related to routing protocol design but nonetheless provided for reference since they do have direct consequences on the security of routing.
At the conceptual level, security within an information system in general and applied to RPL in particular is concerned with the primary issues of authentication, access control, data confidentiality, data integrity, and non-repudiation. In the context of RPL:
It is recognized that, besides those security issues captured in the ISO 7498-2 model, availability, is a security requirement:
It should be emphasized here that for RPL security the above requirements must be complemented by the proper security policies and enforcement mechanisms to ensure that security objectives are met by a given RPL implementation.
The requirements work detailed in Urban Requirements ([RFC5548]), Industrial Requirements ([RFC5673]), Home Automation ([RFC5826], and Building Automation ([RFC5867]) have identified specific issues and constraints of routing in LLNs. The following is a list of observations from those requirements and evaluation of their impact on routing security considerations.
The above list considers how an LLN's physical constraints, size, operations, and variety of application areas may impact security. However, it is the combinations of these factors that particularly stress the security concerns. For instance, securing routing for a large number of autonomous devices that are left in unattended locations with limited physical security presents challenges that are not found in the common circumstance of administered networked routers. The following subsection sets up the security objectives for the routing protocol designed by the ROLL WG.
This subsection applies the ISO 7498-2 model to routing assets and access points, taking into account the LLN issues, to develop a set of RPL security objectives.
Since the fundamental function of a routing protocol is to build routes for forwarding packets, it is essential to ensure that:
In conjunction, it is necessary to be assured that
However, when trust cannot be fully vested through authentication of the principals alone, i.e., concerns of insider attack, assurance of the truthfulness and timeliness of the received routing/topology information is necessary. With regard to confidentiality, protecting the routing/topology information from unauthorized exposure may be desirable in certain cases but is in itself less pertinent in general to the routing function.
One of the main problems of synchronizing security states of sleepy nodes, as listed in the last subsection, lies in difficulties in authentication; these nodes may not have received in time the most recent update of security material. Similarly, the issues of minimal manual configuration, prolonged rollout and delayed addition of nodes, and network topology changes also complicate key management. Hence, routing in LLNs needs to bootstrap the authentication process and allow for flexible expiration scheme of authentication credentials.
The vulnerability brought forth by some special-function nodes, e.g., LBRs, requires the assurance, particularly in a security context,
There are other factors which are not part of RPL but directly affecting its function. These factors include weaker barrier of accessing the data or security material stored on the nodes through physical means; therefore, the internal and external interfaces of a node need to be adequate for guarding the integrity, and possibly the confidentiality, of stored information, as well as the integrity of routing and route generation processes.
Each individual system's use and environment will dictate how the above objectives are applied, including the choices of security services as well as the strengths of the mechanisms that must be implemented. The next two sections take a closer look at how the RPL security objectives may be compromised and how those potential compromises can be countered.
[RFC4593] provides a detailed review of the threat sources: outsiders and byzantine. RPL has the same threat sources.
This section outlines general categories of threats under the ISO 7498-2 model and highlights the specific attacks in each of these categories for RPL. As defined in [RFC4949], a threat is "a potential for violation of security, which exists when there is a circumstance, capability, action, or event that could breach security and cause harm."
An attack is "an assault on system security that derives from an intelligent threat, i.e., an intelligent act that is a deliberate attempt (especially in the sense of a method or technique) to evade security services and violate the security policy of a system."
The subsequent subsections consider the threats and the attacks that can cause security breaches under the ISO 7498-2 model to the routing assets and via the routing points of access identified in Section 4.1. The assessment steps through the security concerns of each routing asset and looks at the attacks that can exploit routing points of access. The threats and attacks identified are based on the routing model analysis and associated review of the existing literature. The source of the attacks is assumed to be from either inside or outside attackers. While some attackers inside the network will be using compromised nodes, and therefore are only able to do what an ordinary node can ("node-equivalent"), other attacks may not limited in memory, CPU, power consumption or long term storage. Moore's law favours the attacker with access to the latest capabilities, while the defenders will remain in place for years to decades.
If an attacker can join a network using any identity, then it may be able to assume the role of a legitimate (and existing node). It may be able to report false readings (in metering applications), or provide inappropriate control messages (in control systems involving actuators) if the security of the application is implied by the security of the routing system.
Even in systems where there application layer security, the ability to impersonate a node would permit an attacker to direct traffic to itself. This may permit various on-path attacks which would otherwise be difficult, such replaying, delaying, or duplicating (application) control messages.
If an attacker can join a network using any identify, then it can pretend to be a legitimate node, receiving any service legitimate nodes receive. It may also be able to report false readings (in metering applications), or provide inappropriate authorizations (in control systems involving actuators), or perform any other attacks that are facilitated by being able to direct traffic towards itself.
If an attacker can join a network with any identify, then it can continously do so with new (random) identities. This act may drain down the resources of the network (battery, RAM, bandwidth). This may cause legitimate nodes of the network to be unable to communicate.
The assessment in Section 4.2 indicates that there are attacks against the confidentiality of routing information at all points of access. This threat may result in disclosure, as described in Section 3.1.2 of [RFC4593], and may involve a disclosure of routing information.
Routing exchanges include both routing information as well as information associated with the establishment and maintenance of neighbor state information. As indicated in Section 4.1, the associated routing information assets may also include device specific resource information, such as available memory, remaining power, etc., that may be metrics of the routing protocol.
The routing exchanges will contain reachability information, which would identify the relative importance of different nodes in the network. Nodes higher up in the DODAG, to which more streams of information flow, would be more interesting targets for other attacks, and routing exchange exposures can identify them.
Routes (which may be maintained in the form of the protocol forwarding table) and neighbor topology information are the products of the routing process that are stored within the node device databases.
The exposure of this information will allow attackers to gain direct access to the configuration and connectivity of the network thereby exposing routing to targeted attacks on key nodes or links. Since routes and neighbor topology information is stored within the node device, attacks on the confidentiality of the information will apply to the physical device including specified and unspecified internal and external interfaces.
The forms of attack that allow unauthorized access or disclosure of the routing information will include:
Both of these attack vectors are considered a device specific issue, and are out of scope for RPL to defend against. In some applications, physical device compromise may be a real threat and it may be necessary to provide for other devices to securely detect a compromised device and react quickly to exclude it.
The assessment in Section 4.2 indicates that information and identity assets are exposed to integrity threats from all points of access. In other words, the integrity threat space is defined by the potential for exploitation introduced by access to assets available through routing exchanges and the on-device storage.
Manipulation of routing information that range from neighbor states to derived routes will allow unauthorized sources to influence the operation and convergence of the routing protocols and ultimately impact the forwarding decisions made in the network.
Manipulation of topology and reachability information will allow unauthorized sources to influence the nodes with which routing information is exchanged and updated. The consequence of manipulating routing exchanges can thus lead to sub-optimality and fragmentation or partitioning of the network by restricting the universe of routers with which associations can be established and maintained.
A sub-optimal network may use too much power and/or may congest some routes leading to premature failure of a node, and a denial of service on the entire network.
In addition, being able to attract network traffic can make a blackhole attack more damaging.
The forms of attack that allow manipulation to compromise the content and validity of routing information include
Falsification or misappropriation of node identity between routing participants opens the door for other attacks; it can also cause incorrect routing relationships to form and/or topologies to emerge. Routing attacks may also be mounted through less sophisticated node identity misappropriation in which the valid information broadcast or exchanged by a node is replayed without modification. The receipt of seemingly valid information that is however no longer current can result in routing disruption, and instability (including failure to converge). Without measures to authenticate the routing participants and to ensure the freshness and validity of the received information the protocol operation can be compromised. The forms of attack that misuse node identity include
The assessment in Section 4.2 indicates that the process and resources assets are exposed to threats against availability; attacks in this category may exploit directly or indirectly information exchange or forwarding (see [RFC4732] for a general discussion).
Interference is the threat action and disruption is threat consequence that allows attackers to influence the operation and convergence of the routing protocols by impeding the routing information exchange.
The forms of attack that allow interference or disruption of routing exchange include:
In addition, attacks may also be directly conducted at the physical layer in the form of jamming or interfering.
The disruption of the network traffic forwarding capability will undermine the central function of network routers and the ability to handle user traffic. This affects the availability of the network because of the potential to impair the primary capability of the network.
In addition to physical layer obstructions, the forms of attack that allows disruption of network traffic forwarding include [Karlof2003]
|Node_1|--(msg1|msg2|msg3)-->|Attacker|--(msg1|msg3)-->|Node_2|
Figure 2: Selective forwarding example
|Node_1|-------------Unreachable---------x|Node_2|
| ^
| Private Link |
'-->|Attacker_1|===========>|Attacker_2|--'
Figure 3: Wormhole Attacks
|Node_1| |Node_4|
| |
`--------. |
Falsify as \ |
Good Link \ | |
To Node_5 \ | |
\ V V
|Node_2|-->|Attacker|--Not Forwarded---x|Node_5|
^ ^ \
| | \ Falsify as
| | \Good Link
/ | To Node_5
,-------' |
| |
|Node_3| |Node_i|
Figure 4: sinkhole attack example
These attacks are generally done to both control plane and forwarding plane traffic. A system that prevents control plane traffic (RPL messages) from being diverted in these ways will also prevent actual data from being diverted.
Attacks mounted against the communication channel resource assets needed by the routing protocol can be used as a means of disrupting its operation. However, while various forms of Denial of Service (DoS) attacks on the underlying transport subsystem will affect routing protocol exchanges and operation (for example physical layer RF jamming in a wireless network or link layer attacks), these attacks cannot be countered by the routing protocol. As such, the threats to the underlying transport network that supports routing is considered beyond the scope of the current document. Nonetheless, attacks on the subsystem will affect routing operation and so must be directly addressed within the underlying subsystem and its implemented protocol layers.
A potential threat consequence can arise from attempts to overload the node resource asset by initiating exchanges that can lead to the exhaustion of processing, memory, or energy resources. The establishment and maintenance of routing neighbors opens the routing process to engagement and potential acceptance of multiple neighboring peers. Association information must be stored for each peer entity and for the wireless network operation provisions made to periodically update and reassess the associations. An introduced proliferation of apparent routing peers can therefore have a negative impact on node resources.
Node resources may also be unduly consumed by attackers attempting uncontrolled topology peering or routing exchanges, routing replays, or the generating of other data traffic floods. Beyond the disruption of communications channel resources, these consequences may be able to exhaust node resources only where the engagements are able to proceed with the peer routing entities. Routing operation and network forwarding functions can thus be adversely impacted by node resources exhaustion that stems from attacks that include:
By recognizing the characteristics of LLNs that may impact routing, this analysis provides the basis for understanding the capabilities within RPL used to deter the identified attacks and mitigate the threats. The following subsections consider such countermeasures by grouping the attacks according to the classification of the ISO 7498-2 model so that associations with the necessary security services are more readily visible.
Attacks to disclosure routing information may be mounted at the level of the routing information assets, at the points of access associated with routing exchanges between nodes, or through device interface access. To gain access to routing/topology information, the attacker may rely on a compromised node that deliberately exposes the information during the routing exchange process, may rely on passive wiretapping or traffic analysis, or may attempt access through a component or device interface of a tampered routing node.
A deliberate exposure attack is one in which an entity that is party to the routing process or topology exchange allows the routing/topology information or generated route information to be exposed to an unauthorized entity.
For instance, due to mis-configuration or inappropriate enabling of a diagnostic interface, an entity might be copying ("bridging") traffic from a secured ESSID/PAN to an unsecured interface.
A prerequisite to countering this attack is to ensure that the communicating nodes are authenticated prior to data encryption applied in the routing exchange. Authentication ensures that the nodes are who they claim to be even though it does not provide an indication of whether the node has been compromised.
To mitigate the risk of deliberate exposure, the process that communicating nodes use to establish session keys must be peer-to-peer (i.e., between the routing initiating and responding nodes). This helps ensure that neither node is exchanging routing information with another peer without the knowledge of both communicating peers. For a deliberate exposure attack to succeed, the comprised node will need to be more overt and take independent actions in order to disclose the routing information to 3rd party.
Note that the same measures which apply to securing routing/topology exchanges between operational nodes must also extend to field tools and other devices used in a deployed network where such devices can be configured to participate in routing exchanges.
A passive wiretap attack seeks to breach routing confidentiality through passive, direct analysis and processing of the information exchanges between nodes.
Passive wiretap attacks can be directly countered through the use of data encryption for all routing exchanges. Only when a validated and authenticated node association is completed will routing exchange be allowed to proceed using established session keys and an agreed encryption algorithm. The mandatory to implement CCM mode AES-128 method, is described in [RFC3610], and is believed to be secure against a brute force attack by even the most well-equipped adversary.
The significant challenge for RPL is in the provisioning of the key, which in some modes of RFC6550 is used network-wide. RFC6550 does not solve this problem, and it is the subject of significant future work: see, for instance: [AceCharterProposal], [SolaceProposal], [SmartObjectSecurityWorkshop].
A number of deployments, such as [ZigBeeIP] specify no layer-3/RPL encryption or authentication and rely upon similiar security at layer-2. These networks are immune to outside wiretapping attacks, but are vulnerable to passive (and active) routing attacks through compromises of nodes. (see Section 8.2).
Section 10.9 of [RFC6550] specifies AES-128 in CCM mode with a 32-bit MAC.
Section 5.6 Zigbee IP [ZigBeeIP] specifies use of CCM, with PANA and EAP-TLS for key management.
Traffic analysis provides an indirect means of subverting confidentiality and gaining access to routing information by allowing an attacker to indirectly map the connectivity or flow patterns (including link-load) of the network from which other attacks can be mounted. The traffic analysis attack on an LLN, especially one founded on shared medium, is passive and relies on the ability to read the immutable source/destination layer-2 and/or layer-3 routing information that must remain unencrypted to permit network routing.
One way in which passive traffic analysis attacks can be muted is through the support of load balancing that allows traffic to a given destination to be sent along diverse routing paths. RPL does not generally support multi-path routing within a single DODAG. Multiple DODAGs are supported in the protocol, and an implementation could make use of that. RPL does not have any inherent or standard way to guarantee that the different DODAGs would have significantly diverse paths. Having the diverse DODAGs routed at different border routers might work in some instances, and this could be combined with a multipath technology like MPTCP ([RFC6824]. It is unlikely that it will be affordable in many LLNs, as few deployments will have memory space for more than a few sets of DODAG tables.
Another approach to countering passive traffic analysis could be for nodes to maintain constant amount of traffic to different destinations through the generation of arbitrary traffic flows; the drawback of course would be the consequent overhead and energy expenditure.
The only means of fully countering a traffic analysis attack is through the use of tunneling (encapsulation) where encryption is applied across the entirety of the original packet source/destination addresses. Deployments which use layer-2 security that includes encryption already do this for all traffic.
Where LLN nodes are deployed in the field, measures are introduced to allow for remote retrieval of routing data and for software or field upgrades. These paths create the potential for a device to be remotely accessed across the network or through a provided field tool. In the case of network management a node can be directly requested to provide routing tables and neighbor information.
To ensure confidentiality of the node routing information against attacks through remote access, any local or remote device requesting routing information must be authenticated, and must be authorized for that access. Since remote access is not invoked as part of a routing protocol, security of routing information stored on the node against remote access will not be addressable as part of the routing protocol.
Integrity attack countermeasures address routing information manipulation, as well as node identity and routing information misuse. Manipulation can occur in the form of falsification attack and physical compromise. To be effective, the following development considers the two aspects of falsification, namely, the unauthorized modifications and the overclaiming and misclaiming content. The countering of physical compromise was considered in the previous section and is not repeated here. With regard to misuse, there are two types of attacks to be deterred, identity attacks and replay attacks.
Unauthorized modifications may occur in the form of altering the message being transferred or the data stored. Therefore, it is necessary to ensure that only authorized nodes can change the portion of the information that is allowed to be mutable, while the integrity of the rest of the information is protected, e.g., through well-studied cryptographic mechanisms.
Unauthorized modifications may also occur in the form of insertion or deletion of messages during protocol changes. Therefore, the protocol needs to ensure the integrity of the sequence of the exchange sequence.
The countermeasure to unauthorized modifications needs to:
Both overclaiming and misclaiming aim to introduce false routes or a false topology that would not occur otherwise, while there are not necessarily unauthorized modifications to the routing messages or information. In order to counter overclaiming, the capability to determine unreasonable routes or topology is required.
The counter to overclaiming and misclaiming may employ:
RPL includes no specific mechanisms in the protocol to counter overclaims or misclaims. An implementation could have specific heuristics implemented locally.
Identity attacks, sometimes simply called spoofing, seek to gain or damage assets whose access is controlled through identity. In routing, an identity attacker can illegitimately participate in routing exchanges, distribute false routing information, or cause an invalid outcome of a routing process.
A perpetrator of Sybil attacks assumes multiple identities. The result is not only an amplification of the damage to routing, but extension to new areas, e.g., where geographic distribution is explicitly or implicitly an asset to an application running on the LLN, for example, the LBR in a P2MP or MP2P LLN.
RPL includes specific public key based authentication at layer-3 that provide for authorization. Many deployments use layer-2 security that includes admission controls at layer-2 using mechanisms such as PANA.
In many routing protocols, message replay can result in false topology and/or routes. This is often counted with some kind of counter to ensure the freshness of the message. Replay of a current, literal RPL message are in general idempotent to the topology. An older (lower DODAGVersionNumber) message, if replayed would be rejected as being stale. The trickle algorithm further dampens the effect of any such replay, as if the message was current, then it would contain the same information as before, and it would cause no network changes.
Replays may well occur in some radio technologies (not very likely, 802.15.4) as a result of echos or reflections, and so some replays must be assumed to occur naturally.
Note that for there to be no affect at all, the replay must be done with the same apparent power for all nodes receiving the replay. A change in apparent power might change the metrics through changes to the ETX and therefore might affect the routing even though the contents of the packet were never changed. Any replay which appears to be different should be analyzed as a Selective Forwarding Attack, Sinkhole Attack or Wormhole Attack.
Where a node is captured or compromised but continues to operate for a period with valid network security credentials, the potential exists for routing information to be manipulated. This compromise of the routing information could thus exist in spite of security countermeasures that operate between the peer routing devices.
Consistent with the end-to-end principle of communications, such an attack can only be fully addressed through measures operating directly between the routing entities themselves or by means of external entities able to access and independently analyze the routing information. Verification of the authenticity and liveliness of the routing entities can therefore only provide a limited counter against internal (Byzantine) node attacks.
For link state routing protocols where information is flooded with, for example, areas (OSPF [RFC2328]) or levels (ISIS [RFC1142]), countermeasures can be directly applied by the routing entities through the processing and comparison of link state information received from different peers. By comparing the link information from multiple sources decisions can be made by a routing node or external entity with regard to routing information validity; see Chapter 2 of [Perlman1988] for a discussion on flooding attacks.
For distance vector protocols, such as RPL, where information is aggregated at each routing node it is not possible for nodes to directly detect Byzantine information manipulation attacks from the routing information exchange. In such cases, the routing protocol must include and support indirect communications exchanges between non-adjacent routing peers to provide a secondary channel for performing routing information validation. S-RIP [Wan2004] is an example of the implementation of this type of dedicated routing protocol security where the correctness of aggregate distance vector information can only be validated by initiating confirmation exchanges directly between nodes that are not routing neighbors.
RPL does not provide any direct mechanisms like S-RIP. It does listen to multiple parents, and may switch parents if it begins to suspect that it is being lied to.
As alluded to before, availability requires that routing information exchanges and forwarding mechanisms be available when needed so as to guarantee proper functioning of the network. This may, e.g., include the correct operation of routing information and neighbor state information exchanges, among others. We will highlight the key features of the security threats along with typical countermeasures to prevent or at least mitigate them. We will also note that an availability attack may be facilitated by an identity attack as well as a replay attack, as was addressed in Section 7.2.3 and Section 7.2.4, respectively.
HELLO Flood [Karlof2003],[I-D.suhopark-hello-wsn] and ACK Spoofing attacks are different but highly related forms of attacking an LLN. They essentially lead nodes to believe that suitable routes are available even though they are not and hence constitute a serious availability attack.
A HELLO attack mounted against RPL would involve sending out (or replaying) DIO messages by the attacker. Lower power LLN nodes might then attempt to join the DODAG at a lower rank than they would otherwise.
The most effective method from [I-D.suhopark-hello-wsn] is the verify bidirectionality. A number of layer-2 links are arranged in controller/spoke arrangements, and continuously are validating connectivity at layer 2.
In addition, in order to calculate metrics, the ETX must be computed, and this involves, in general, sending a number of messages between nodes which are believed to be adjacent. [I-D.kelsey-intarea-mesh-link-establishment] is one such protocol.
In order to join the DODAG, a DAO message is sent upwards. In RPL the DAO is acknowledged by the DAO-ACK message. This clearly checks bidirectionality at the control plane.
As discussed in section 5.1, [I-D.suhopark-hello-wsn] a receiver with a sensitive receiver could well hear the DAOs, and even send DAO-ACKs as well. Such a node is a form of wormhole attack.
These attacks are also all easily defended against using either layer-2 or layer-3 authentication. Such an attack could only be made against a completely open network (such as might be used for provisioning new nodes), or by a compromised node.
Overload attacks are a form of DoS attack in that a malicious node overloads the network with irrelevant traffic, thereby draining the nodes' energy store more quickly, when the nodes rely on batteries or energy scavenging. It thus significantly shortens the lifetime of networks of energy-constrained nodes and constitutes another serious availability attack.
With energy being one of the most precious assets of LLNs, targeting its availability is a fairly obvious attack. Another way of depleting the energy of an LLN node is to have the malicious node overload the network with irrelevant traffic. This impacts availability since certain routes get congested which:
Overload attacks can be countered by deploying a series of mutually non-exclusive security measures:
As for the first one, a simple approach to minimize the harmful impact of an overload attack is to introduce traffic quotas. This prevents a malicious node from injecting a large amount of traffic into the network, even though it does not prevent said node from injecting irrelevant traffic at all. Another method is to isolate nodes from the network at the network layer once it has been detected that more traffic is injected into the network than allowed by a prior set or dynamically adjusted threshold. Finally, if communication is sufficiently secured, only trusted nodes can receive and forward traffic which also lowers the risk of an overload attack.
Receiving nodes that validate signatures and sending nodes that encrypt messages need to be cautious of cryptographic processing usage when validating signatures and encrypting messages. Where feasible, certificates should be validated prior to use of the associated keys to counter potential resource overloading attacks. The associated design decision needs to also consider that the validation process requires resources and thus itself could be exploited for attacks. Alternatively, resource management limits can be placed on routing security processing events (see the comment in Section 6, paragraph 4, of [RFC5751]).
Selective forwarding attacks are a form of DoS attack which impacts the availability of the generated routing paths.
A selective forwarding attack may be done by a node involved with the routing process, or it may be done by what otherwise appears to be a passive antenna or other RF feature or device, but is in fact an active (and selective) device. An RF antenna/repeater which is not selective, is not a threat.
An insider malicious node basically blends neatly in with the network but then may decide to forward and/or manipulate certain packets. If all packets are dropped, then this attacker is also often referred to as a "black hole". Such a form of attack is particularly dangerous if coupled with sinkhole attacks since inherently a large amount of traffic is attracted to the malicious node and thereby causing significant damage. In a shared medium, an outside malicious node would selectively jam overheard data flows, where the thus caused collisions incur selective forwarding.
Selective Forwarding attacks can be countered by deploying a series of mutually non-exclusive security measures:
The first measure basically guarantees that if a message gets lost on a particular routing path due to a malicious selective forwarding attack, there will be another route which successfully delivers the data. Such a method is inherently suboptimal from an energy consumption point of view; it is also suboptimal from a network utilization perspective. The second method basically involves a constantly changing routing topology in that next-hop routers are chosen from a dynamic set in the hope that the number of malicious nodes in this set is negligible. A routing protocol that allows for disjoint routing paths may also be useful.
In sinkhole attacks, the malicious node manages to attract a lot of traffic mainly by advertising the availability of high-quality links even though there are none [Karlof2003]. It hence constitutes a serious attack on availability.
The malicious node creates a sinkhole by attracting a large amount of, if not all, traffic from surrounding neighbors by advertising in and outwards links of superior quality. Affected nodes hence eagerly route their traffic via the malicious node which, if coupled with other attacks such as selective forwarding, may lead to serious availability and security breaches. Such an attack can only be executed by an inside malicious node and is generally very difficult to detect. An ongoing attack has a profound impact on the network topology and essentially becomes a problem of flow control.
Sinkhole attacks can be countered by deploying a series of mutually non-exclusive security measures:
Some LLNs may provide for geolocation services, often derived from solving triangulation equations from radio delay calculations, such calculations could in theory be subverted by a sinkhole that transmitted at precisely the right power in a node to node fashion.
While geographic knowledge could help assure that traffic always went in the physical direction desired, it would not assure that the traffic was taking the most efficient route, as the lowest cost real route might be match the physical topology; such as when different parts of an LLN are connected by high-speed wired networks.
In wormhole attacks at least two malicious nodes claim to have a short path between themselves [Karlof2003]. This changes the availability of certain routing paths and hence constitutes a serious security breach.
Essentially, two malicious insider nodes use another, more powerful, transmitter to communicate with each other and thereby distort the would-be-agreed routing path. This distortion could involve shortcutting and hence paralyzing a large part of the network; it could also involve tunneling the information to another region of the network where there are, e.g., more malicious nodes available to aid the intrusion or where messages are replayed, etc.
In conjunction with selective forwarding, wormhole attacks can create race conditions which impact topology maintenance, routing protocols as well as any security suits built on "time of check" and "time of use".
A pure wormhole attack is nearly impossible to detect. A wormhole which is used in order to subsequently mount another kind of attack would be defeated by defeating the other attack. A perfect wormhole, in which there is nothing adverse that occurs to the traffic, would be difficult to call an attack. The worst thing that a benign wormhole can do in such a situation is to cease to operate (become unstable), causing the network to have to recalculate routes.
A highly unstable wormhole is no different than a radio opaque (i.e. metal) door that opens and closes a lot. RPL includes hysteresis in its objective functions [RFC6719] in an attempt to deal with frequent changes to the ETX between nodes.
The assessments and analysis in Section 6 examined all areas of threats and attacks that could impact routing, and the countermeasures presented in Section 7 were reached without confining the consideration to means only available to routing. This section puts the results into perspective; dealing with those threats which are endemic to this field, those which have been mitigated through RPL protocol design, and those which require specific decisions to be made as part of provisioning a network.
The first part of this section, Section 8.1 to Section 8.3, is a description of RPL security features that address specific threats. The second part of this section, Section 8.4, discusses issues of provisioning of security aspects that may impact routing but that also require considerations beyond the routing protocol, as well as potential approaches.
RPL employs multicast and so these alternative communications modes MUST be secured with the same routing security services specified in this section. Furthermore, irrespective of the modes of communication, nodes MUST provide adequate physical tamper resistance commensurate with the particular application domain environment to ensure the confidentiality, integrity, and availability of stored routing information.
With regard to confidentiality, protecting the routing/topology information from unauthorized disclosure is not directly essential to maintaining the routing function. Breaches of confidentiality may lead to other attacks or the focusing of an attacker's resources (see Section 6.2) but does not of itself directly undermine the operation of the routing function. However, to protect against, and reduce consequences from other more direct attacks, routing information should be protected. Thus, to secure RPL:
Where confidentiality is incorporated into the routing exchanges, encryption algorithms and key lengths need to be specified in accordance with the level of protection dictated by the routing protocol and the associated application domain transport network. For most networks, this means use of AES128 in CCM mode, but this needs to be specified clearly in the applicability statement.
In terms of the life time of the keys, the opportunity to periodically change the encryption key increases the offered level of security for any given implementation. However, where strong cryptography is employed, physical, procedural, and logical data access protection considerations may have more significant impact on cryptoperiod selection than algorithm and key size factors. Nevertheless, in general, shorter cryptoperiods, during which a single key is applied, will enhance security.
Given the mandatory protocol requirement to implement routing node authentication as part of routing integrity (see Section 8.2), key exchanges may be coordinated as part of the integrity verification process. This provides an opportunity to increase the frequency of key exchange and shorten the cryptoperiod as a complement to the key length and encryption algorithm required for a given application domain.
The integrity of routing information provides the basis for ensuring that the function of the routing protocol is achieved and maintained. To protect integrity, RPL must either run using only the Secure versions of the messages, or must run over a layer-2 that uses channel binding between node identity and transmissions.
Some layer-2 security mechanisms use a single key for the entire network, and these networks can not provide significant amount of integrity protection, as any node that has that key may impersonate any other node. This mode of operation is likely acceptable when an entire deployment is under the control of a single administrative entity.
Other layer-2 security mechanisms form a unique session key for every pair of nodes that needs to communicate; this is often called a per-link key. Such networks can provide a strong degree of origin authentication and integrity on unicast messages.
However, some RPL messages are broadcast, and even when per-node layer-2 security mechanisms are used, the integrity and origin authentication of broadcast messages can not be as trusted due to the proliferation of the key used to secure them.
RPL has two specific options which are broadcast in RPL Control Messages: the DODAG Information Object (DIO), and the DODAG Information Solicitation (DIS). The purpose of the DIS is to cause potential parents to reply with a DIO, so the integrity of the DIS is not of great concern. The DIS may also be unicast.
The DIO is a critical piece of routing and carries many critical parameters. RPL provides for asymmetric authentication at layer 3 of the RPL Control Message carrying the DIO and this may be warranted in some deployments. A node could, if it felt that the DIO that it had received was suspicious, send a unicast DIS message to the node in question, and that node would reply with a unicast DIS. Those messages could be protected with the per-link key.
Availability of routing information is linked to system and network availability which in the case of LLNs require a broader security view beyond the requirements of the routing entities. Where availability of the network is compromised, routing information availability will be accordingly affected. However, to specifically assist in protecting routing availability, nodes:
The functioning of the routing security services requires keys and credentials. Therefore, even though not directly a RPL security requirement, an LLN MUST have a process for initial key and credential configuration, as well as secure storage within the associated devices. Anti-tampering SHOULD be a consideration in physical design. Beyond initial credential configuration, an LLN is also encouraged to have automatic procedures for the revocation and replacement of the maintained security credentials.
While RPL has secure modes, but some modes are impractical without use of public key cryptography believed to be too expensive by many. RPL layer-3 security will often depend upon existing LLN layer-2 security mechanisms, which provides for node authentication, but little in the way of node authorization.
This memo includes no request to IANA.
The analysis presented in this document provides security analysis and design guidelines with a scope limited to RPL. Security services are identified as requirements for securing RPL. The specific mechanisms to be used to deal with each threat is specified in link-layer and deployment specific applicability statements.
The authors would like to acknowledge the review and comments from Rene Struik and JP Vasseur. The authors would also like to acknowledge the guidance and input provided by the RPL Chairs, David Culler, and JP Vasseur, and the Area Director Adrian Farrel.
This document started out as a combined threat and solutions document. As a result of security review, the document was split up by RPL co-Chair Michael Richardson and security Area Director Sean Turner as it went through the IETF publication process. The solutions to the threats are application and layer-2 specific, and have therefore been moved to the relevant applicability statements.
Ines Robles and Robert Cragie kept track of the many issues that were raised during the development of this document
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. |
| [RFC4107] | Bellovin, S. and R. Housley, "Guidelines for Cryptographic Key Management", BCP 107, RFC 4107, June 2005. |
| [RFC4301] | Kent, S. and K. Seo, "Security Architecture for the Internet Protocol", RFC 4301, December 2005. |
| [RFC6550] | Winter, T., Thubert, P., Brandt, A., Hui, J., Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, JP. and R. Alexander, "RPL: IPv6 Routing Protocol for Low-Power and Lossy Networks", RFC 6550, March 2012. |
| [RFC6719] | Gnawali, O. and P. Levis, "The Minimum Rank with Hysteresis Objective Function", RFC 6719, September 2012. |
| [RFC7102] | Vasseur, JP., "Terms Used in Routing for Low-Power and Lossy Networks", RFC 7102, January 2014. |
| [ZigBeeIP] | ZigBee Public Document 15-002r00, "ZigBee IP Specification", 2013. |