Internet Engineering Task Force (IETF) H. Moustafa Request for Comments: 5713 France Telecom Category: Informational H. Tschofenig ISSN: 2070-1721 Nokia Siemens Networks S. De Cnodder Alcatel-Lucent January 2010
Security Threats and Security Requirements for the Access Node Control Protocol (ANCP)
Abstract
The Access Node Control Protocol (ANCP) aims to communicate Quality of Service (QoS)-related, service-related, and subscriber-related configurations and operations between a Network Access Server (NAS) and an Access Node (e.g., a Digital Subscriber Line Access Multiplexer (DSLAM)). The main goal of this protocol is to allow the NAS to configure, manage, and control access equipment, including the ability for the Access Nodes to report information to the NAS.
This present document investigates security threats that all ANCP nodes could encounter. This document develops a threat model for ANCP security, with the aim of deciding which security functions are required. Based on this, security requirements regarding the Access Node Control Protocol are defined.
Status of This Memo
This document is not an Internet Standards Track specification; it is published for informational purposes.
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Not all documents approved by the IESG are a candidate for any level of Internet Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc5713.
Moustafa, et al. Informational [Page 1]
RFC 5713 ANCP Threats January 2010
Copyright Notice
Copyright (c) 2010 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
The Access Node Control Protocol (ANCP) aims to communicate QoS- related, service-related, and subscriber-related configurations and operations between a Network Access Server (NAS) and an Access Node (e.g., a Digital Subscriber Line Access Multiplexer (DSLAM)).
[ANCP-FRAME] illustrates the framework, usage scenarios, and general requirements for ANCP. This document focuses on describing security threats and deriving security requirements for the Access Node Control Protocol, considering the ANCP use cases defined in [ANCP-FRAME] as well as the guidelines for IETF protocols' security requirements given in [RFC3365]. Section 5 and Section 6, respectively, describe the potential attacks and the different attack forms that are liable to take place within ANCP, while Section 7 applies the described potential attacks to ANCP and its different use cases. Security policy negotiation, including authentication and authorization to define the per-subscriber policy at the policy/AAA (Authentication, Authorization, and Accounting) server, is out of the scope of this work. As a high-level summary, the following aspects need to be considered:
Message Protection:
Signaling message content can be protected against eavesdropping, modification, injection, and replay while in transit. This applies to both ANCP headers and payloads.
Prevention against Impersonation:
It is important that protection be available against a device impersonating an ANCP node (i.e., an unauthorized device generating an ANCP message and pretending it was generated by a valid ANCP node).
Prevention of Denial-of-Service Attacks:
ANCP nodes and the network have finite resources (state storage, processing power, bandwidth). It is important to protect against exhaustion attacks on these resources and to prevent ANCP nodes from being used to launch attacks on other network elements.
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], with the
Moustafa, et al. Informational [Page 3]
RFC 5713 ANCP Threats January 2010
qualification that, unless otherwise stated, they apply to the design of the Access Node Control Protocol (ANCP), not its implementation or application.
The relevant components are described in Section 3.
As described in [ANCP-FRAME] and schematically shown in Figure 1, the Access Node Control system consists of the following components:
Network Access Server (NAS):
A NAS provides access to a service (e.g., network access) and operates as a client of the AAA protocol. The AAA client is responsible for passing authentication information to designated AAA servers and then acting on the response that is returned.
Authentication, Authorization, and Accounting (AAA) server:
A AAA server is responsible for authenticating users, authorizing access to services, and returning authorization information (including configuration parameters) back to the AAA client to deliver service to the user. As a consequence, service usage accounting might be enabled and information about the user's resource usage will be sent to the AAA server.
Access Node (AN):
The AN is a network device, usually located at a service provider central office or street cabinet, that terminates access-loop connections from subscribers. In case the access loop is a Digital Subscriber Line (DSL), this is often referred to as a DSL Access Multiplexer (DSLAM).
Customer Premises Equipment (CPE):
A CPE is a device located inside a subscriber's premise that is connected at the LAN side of the Home Gateway (HGW).
Home Gateway (HGW):
The HGW connects the different Customer Premises Equipments (CPEs) to the Access Node and the access network. In case of DSL, the HGW is a DSL Network Termination (NT) that could either operate as a layer 2 bridge or as a layer 3 router. In the latter case, such a device is also referred to as a Routing Gateway (RG).
Moustafa, et al. Informational [Page 4]
RFC 5713 ANCP Threats January 2010
Aggregation Network:
The aggregation network provides traffic aggregation from multiple ANs towards the NAS. ATM or Ethernet transport technologies can be used.
For the threat analysis, this document focuses on the ANCP communication between the Access Node and the NAS. However, communications with the other components (such as HGW, CPE, and the AAA server) play a role in the understanding of the system architecture and of what triggers ANCP communications. Note that the NAS and the AN might belong to two different administrative realms. The threat model and the security requirements in this document consider this latter case.
In the absence of an attack, the NAS receives configuration information from the AAA server related to a CPE attempting to access the network. A number of parameters, including Quality of Service information, need to be conveyed to the Access Node in order to become effective. The Access Node Control Protocol is executed between the NAS and the AN to initiate control requests. The AN returns responses to these control requests and provides information reports.
For this to happen, the following individual steps must occur:
o The AN discovers the NAS.
o The AN needs to start the protocol communication with the NAS to announce its presence.
Moustafa, et al. Informational [Page 5]
RFC 5713 ANCP Threats January 2010
o The AN and the NAS perform a capability exchange.
o The NAS sends requests to the AN.
o The AN processes these requests, authorizes the actions, and responds with the appropriate answer. In order to fulfill the commands, it might be necessary for the AN to communicate with the HGW or other nodes, for example, as part of a keep-alive mechanism.
o The AN provides status reports to the NAS.
Attackers can be:
o off-path, i.e., they cannot see the messages exchanged between the AN and the NAS;
o on-path, i.e., they can see the messages exchanged between the AN and the NAS.
Both off-path and on-path attackers can be:
o passive, i.e., they do not participate in the network operation but rather listen to all transfers to obtain the maximum possible information;
o active, i.e., they participate in the network operation and can inject falsified packets.
We assume the following threat model:
o An off-path adversary located at the CPE or the HGW.
o An off-path adversary located on the Internet or a regional network that connects one or more NASes and associated access networks to Network Service Providers (NSPs) and Application Service Providers (ASPs).
o An on-path adversary located at network elements between the AN and the NAS.
o An on-path adversary taking control over the NAS.
o An on-path adversary taking control over the AN.
Attackers may direct their efforts either against an individual entity or against a large portion of the access network. Attacks fall into three classes:
o Attacks to disrupt the communication for individual customers.
o Attacks to disrupt the communication of a large fraction of customers in an access network. These also include attacks to the network itself or a portion of it, such as attacks to disrupt the network services or attacks to destruct the network functioning.
o Attacks to gain profit for the attacker through modifying the QoS settings. Also, through replaying old packets (of another privileged client, for instance), an attacker can attempt to configure a better QoS profile on its own DSL line, increasing its own benefit.
A number of denial-of-service (DoS) attacks can cause ANCP nodes to malfunction. When state is established or certain functions are performed without requiring prior authorization, there is a chance to mount denial-of-service attacks. An adversary can utilize this fact to transmit a large number of signaling messages to allocate state at nodes and to cause consumption of resources. Also, an adversary, through DoS, can prevent certain subscribers from accessing certain services. Moreover, DoS can take place at the AN or the NAS themselves, where it is possible for the NAS (or the AN) to intentionally ignore the requests received from the AN (or the NAS) through not replying to them. This causes the sender of the request to retransmit the request, which might allocate additional state at the sender side to process the reply. Allocating more state may result in memory depletion.
Adversaries gaining illegitimate access on the transferred messages can act on these messages, causing integrity violation. Integrity violation can cause unexpected network behavior, leading to a disturbance in the network services as well as in the network functioning.
Protocols may be useful in a variety of scenarios with different security and functional requirements. Different parts of a network (e.g., within a building, across a public carrier's network, or over a private microwave link) may need different levels of protection. It is often difficult to meet these (sometimes conflicting) requirements with a single mechanism or fixed set of parameters; thus, often a selection of mechanisms and parameters is offered. A protocol is required to agree on certain (security) mechanisms and parameters. An insecure parameter exchange or security negotiation protocol can give an adversary the opportunity to mount a downgrading attack to force selection of mechanisms weaker than those mutually desired. Thus, without binding the negotiation process to the legitimate parties and protecting it, ANCP might only be as secure as the weakest mechanism provided (e.g., weak authentication) and the benefits of defining configuration parameters and a negotiation protocol are lost.
An adversary can be placed at the NAS, the AN, or any other network element capturing all traversing packets. Adversaries can thus have unauthorized information access. As well, they can gather information relevant to the network and then use this information in gaining later unauthorized access. This attack can also help adversaries in other malicious purposes -- for example, capturing messages sent from the AN to the NAS announcing that a DSL line is up and containing some information related to the connected client. This could be any form of information about the client and could also be an indicator of whether or not the DSL subscriber is at home at a particular moment.
Since the ANCP sessions are configured in the AN and not in the NAS [ANCP-FRAME], most configurations of ANCP are done in the AN. Consequently, the management attacks to ANCP mainly concern the AN configuration phase. In this context, the AN MIB module could create disclosure- and misconfiguration-related attacks. [ANCP-MIB] defines
Moustafa, et al. Informational [Page 8]
RFC 5713 ANCP Threats January 2010
the vulnerabilities on the management objects within the AN MIB module. These attacks mainly concern the unauthorized changes of the management objects, leading to a number of attacks such as session deletion, a session using an undesired/unsupported protocol, disabling certain ANCP capabilities or enabling undesired capabilities, ANCP packets being sent out to the wrong interface (and thus being received by an unintended receiver), harming the synchronization between the AN and the NAS, and impacting traffic in the network other than ANCP.
The attacks mentioned above in Section 5 can be carried out through the following means:
Message Replay:
This threat scenario covers the case in which an adversary eavesdrops, collects signaling messages, and replays them at a later time (or at a different place or in a different way; e.g., cut-and-paste attacks). Through replaying signaling messages, an adversary might mount denial-of-service and theft-of-service attacks.
Faked Message Injection:
An adversary may be able to inject false error or response messages, causing unexpected protocol behavior and succeeding with a DoS attack. This could be achieved at the signaling-protocol level, at the level of specific signaling parameters (e.g., QoS information), or at the transport layer. An adversary might, for example, inject a signaling message to request allocation of QoS resources. As a consequence, other users' traffic might be impacted. The discovery protocol, especially, exhibits vulnerabilities with regard to this threat scenario.
Messages Modification:
This involves integrity violation, where an adversary can modify signaling messages in order to cause unexpected network behavior. Possible related actions an adversary might consider for its attack are the reordering and delaying of messages, causing a protocol's process failure.
Moustafa, et al. Informational [Page 9]
RFC 5713 ANCP Threats January 2010
Man-in-the-Middle:
An adversary might claim to be a NAS or an AN, acting as a man-in- the-middle to later cause communication and services disruption. The consequence can range from DoS to fraud. An adversary acting as a man-in-the-middle could modify the intercepted messages, causing integrity violation, or could drop or truncate the intercepted messages, causing DoS and a protocol's process failure. In addition, a man-in-the-middle adversary can signal information to an illegitimate entity in place of the right destination. In this case, the protocol could appear to continue working correctly. This may result in an AN contacting a wrong NAS. For the AN, this could mean that the protocol failed for unknown reasons. A man-in-the-middle adversary can also cause downgrading attacks through initiating faked configuration parameters and through forcing selection of weak security parameters or mechanisms.
Eavesdropping:
This is related to adversaries that are able to eavesdrop on transferred messages. The collection of the transferred packets by an adversary may allow traffic analysis or be used later to mount replay attacks. The eavesdropper might learn QoS parameters, communication patterns, policy rules for firewall traversal, policy information, application identifiers, user identities, NAT bindings, authorization objects, network configuration, performance information, and more.
ANCP is susceptible to security threats, causing disruption/ unauthorized access to network services, manipulation of the transferred data, and interference with network functions. Based on the threat model given in Section 3 and the potential attacks presented in Section 5, this section describes the possible attacks against ANCP, considering the four use cases defined in [ANCP-FRAME].
Although ANCP is not involved in the communication between the NAS and the AAA/policy server, the secure communication between the NAS and the AAA/policy server is important for ANCP security. Consequently, this document considers the attacks that are related to the ANCP operation associated with the communication between the NAS and the AAA/Policy server. In other words, the threat model and security requirements in this document take into consideration the data transfer between the NAS and the AAA server, when this data is used within the ANCP operation.
Moustafa, et al. Informational [Page 10]
RFC 5713 ANCP Threats January 2010
Besides the attacks against the four ANCP use cases described in the following subsections, ANCP is susceptible to a number of attacks that can take place during the protocol-establishment phase. These attacks are mainly on-path attacks, taking the form of DoS or man-in- the-middle attacks, which could be as follows:
o Attacks during the session initiation from the AN to the NAS: DoS attacks could take place affecting the session-establishment process. Also, man-in-the-middle attacks could take place, causing message truncation or message modification and leading to session-establishment failure.
o Attacks during the peering establishment: DoS attacks could take place during state synchronization between the AN and the NAS. Also, man-in-the-middle attacks could take place through message modification during identity discovery, which may lead to loss of contact between the AN and the NAS.
o Attacks during capabilities negotiation: Message replay could take place, leading to DoS. Also, man-in- the-middle attacks could take place, leading to message modification, message truncation, or downgrading through advertising lesser capabilities.
This use case concerns the communication of access-loop attributes for dynamic, access-line topology discovery. Since the access-loop rate may change over time, advertisement is beneficial to the NAS to gain knowledge about the topology of the access network for QoS scheduling. Besides data rates and access-loop links identification, other information may also be transferred from the AN to the NAS (examples in case of a DSL access loop are DSL type, maximum achievable data rate, and maximum data rate configured for the access loop). This use case is thus vulnerable to a number of on-path and off-path attacks that can be either active or passive.
On-path attacks can take place between the AN and the NAS, on the AN or on the NAS, during the access-loop attributes transfer. These attacks may be:
o Active, acting on the transferred attributes and injecting falsified packets. The main attacks here are:
* Man-in-the-middle attacks can cause access-loop attributes transfer between the AN and a forged NAS or a forged AN and the NAS, which can directly cause faked attributes and message modification or truncation.
Moustafa, et al. Informational [Page 11]
RFC 5713 ANCP Threats January 2010
* Signaling replay, by an attacker between the AN and the NAS, on the AN or on the NAS itself, causing DoS.
* An adversary acting as man-in-the-middle can cause downgrading through changing the actual data rate of the access loop, which impacts the downstream shaping from the NAS.
o Passive, only learning these attributes. The main attacks here are caused by:
* Eavesdropping through learning access-loop attributes and information about the clients' connection state, and thus impacting their privacy protection.
* Traffic analysis allowing unauthorized information access, which could allow later unauthorized access to the NAS.
Off-path attacks can take place on the Internet, affecting the access-loop attribute sharing between the NAS and the AAA/policy server. These attacks may be:
o Active attacks, which are mainly concerning:
* DoS through flooding the communication links to the AAA/policy server, causing service disruption.
* Man-in-the-middle, causing access-loop configuration retrieval by an illegitimate NAS.
o Passive attacks, gaining information on the access-loop attributes. The main attacks in this case are:
* Eavesdropping through learning access-loop attributes and learning information about the clients' connection states, and thus impacting their privacy protection.
* Traffic analysis allowing unauthorized information access, which could allow later unauthorized access to the NAS.
This use case concerns the dynamic, local-loop line configuration through allowing the NAS to change the access-loop parameters (e.g., rate) in a dynamic fashion. This allows for centralized, subscriber- related service data. This dynamic configuration can be achieved, for instance, through profiles that are pre-configured on ANs. This use case is vulnerable to a number of on-path and off-path attacks.
Moustafa, et al. Informational [Page 12]
RFC 5713 ANCP Threats January 2010
On-path attacks can take place where the attacker is between the AN and the NAS, is on the AN, or is on the NAS. These can be as follows:
o Active attacks, taking the following forms:
* DoS attacks of the AN can take place by an attacker, through replaying the Configure Request messages.
* An attacker on the AN can prevent the AN from reacting on the NAS request for the access-loop configuration, leading to the NAS continually sending the Configure Request message and, hence, allocating additional states.
* Damaging clients' profiles at ANs can take place by adversaries that gained control on the network through discovery of users' information from a previous traffic analysis.
* An adversary can replay old packets, modify messages, or inject faked messages. Such adversary can also be a man-in-the- middle. These attack forms can be related to a privileged client profile (having more services) in order to configure this profile on the adversary's own DSL line, which is less privileged. In order that the attacker does not expose its identity, he may also use these attack forms related to the privileged client profile to configure a number of illegitimate DSL lines. The adversary can also force configuration parameters other than the selected ones, leading to, for instance, downgrading the service for a privileged client.
o Passive attacks, where the attacker listens to the ANCP messages. This can take place as follows:
* Learning configuration attributes is possible during the update of the access-loop configuration. An adversary might profit to see the configuration that someone else gets (e.g., one ISP might be interested to know what the customers of another ISP get and therefore might break into the AN to see this).
Off-path attacks can take place as follows:
o An off-path passive adversary on the Internet can exert eavesdropping during the access-loop configuration retrieval by the NAS from the AAA/policy server.
Moustafa, et al. Informational [Page 13]
RFC 5713 ANCP Threats January 2010
o An off-path active adversary on the Internet can threaten the centralized subscribers-related service data in the AAA/policy server through, for instance, making subscribers' records inaccessible.
In this use case, the NAS can carry out a Remote Connectivity Test using ANCP to initiate an access-loop test between the AN and the HGW. Thus, multiple access-loop technologies can be supported. This use case is vulnerable to a number of active attacks. Most of the attacks in this use case concern the network operation.
On-path active attacks can take place in the following forms:
o Man-in-the-middle attack during the NAS's triggering to the AN to carry out the test, where an adversary can inject falsified signals or can truncate the triggering.
o Message modification can take place during the Subscriber Response message transfer from the AN to the NAS announcing the test results, causing failure of the test operation.
o An adversary on the AN can prevent the AN from sending the Subscriber Response message to the NAS announcing the test results, and hence the NAS will continue triggering the AN to carry out the test, which results in more state being allocated at the NAS. This may result in unavailability of the NAS to the ANs.
Off-path active attacks can take place as follows:
o An adversary can cause DoS during the access-loop test, in case of an ATM-based access loop, when the AN generates loopback cells. This can take place through signal replaying.
o Message truncating can take place by an adversary during the access-loop test, which can lead to service disruption due to assumption of test failures.
In this use case, ANCP could be used in exchanging information between the AN and the NAS, allowing the AN to perform replication inline with the policy and configuration of the subscriber. Also, this allows the NAS to follow subscribers' multicast (source, group) membership and control replication performed by the AN. Four multicast use cases are expected to take place, making use of ANCP; these are typically multicast conditional access, multicast admission
Moustafa, et al. Informational [Page 14]
RFC 5713 ANCP Threats January 2010
control, multicast accounting, and spontaneous admission response. This section gives a high-level description of the possible attacks that can take place in these cases. Attacks that can occur are mostly active attacks.
On-path active attacks can be as follows:
o DoS attacks, causing inability for certain subscribers to access particular multicast streams or only access the multicast stream at a reduced bandwidth, impacting the quality of the possible video stream. This can take place through message replay by an attacker between the AN and the NAS, on the AN or on the NAS. Such DoS attacks can also be done by tempering, for instance, with white/black list configuration or by placing attacks to the bandwidth-admission-control mechanism.
o An adversary on the NAS can prevent the NAS from reacting on the AN requests for white/black/grey lists or for admission control for the access line. The AN in this case would not receive a reply and would continue sending its requests, resulting in more states being allocated at the AN. A similar case happens for admission control when the NAS can also send requests to the AN. When the NAS does not receive a response, it could also retransmit requests, resulting in more state being allocated at the NAS side to process responses. This may result in the unavailability of the NAS to the ANs.
o Man-in-the-middle, causing the exchange of messages between the AN and a forged NAS or a forged AN and the NAS. This can lead to the following:
* Message modification, which can cause service downgrading for legitimate subscribers -- for instance, an illegitimate change of a subscriber's policy.
* Message truncation between the AN and the NAS, which can result in the non-continuity of services.
* Message replay between the AN and the NAS, on the AN or on the NAS, leading to a DoS or services fraud.
* Message modification to temper with accounting information, for example, in order to avoid service charges or, conversely, in order to artificially increase service charges on other users.
Moustafa, et al. Informational [Page 15]
RFC 5713 ANCP Threats January 2010
An off-path active attack is as follows:
o DoS could take place through message replay of join/leave requests by the HGW or CPE, frequently triggering the ANCP activity between the AN and the NAS. DoS could also result from generating heaps of IGMP join/leaves by the HGW or CPE, leading to very high rate of ANCP query/response.
This section presents a number of requirements motivated by the different types of attacks defined in the previous section. These requirements are as follows:
o The protocol solution MUST offer authentication of the AN to the NAS.
o The protocol solution MUST offer authentication of the NAS to the AN.
o The protocol solution MUST allow authorization to take place at the NAS and the AN.
o The protocol solution MUST offer replay protection.
o The protocol solution MUST provide data-origin authentication.
o The protocol solution MUST be robust against denial-of-service (DoS) attacks. In this context, the protocol solution MUST consider a specific mechanism for the DoS that the user might create by sending many IGMP messages.
o The protocol solution SHOULD offer confidentiality protection.
o The protocol solution SHOULD ensure that operations in default configuration guarantees a low number of AN/NAS protocol interactions.
o The protocol solution SHOULD ensure the access control of the management objects and possibly encrypt the values of these objects when sending them over the networks.
This document focuses on security threats, deriving a threat model for ANCP and presenting the security requirements to be considered for the design of ANCP.
Many thanks go to Francois Le Faucher for reviewing this document and for all his useful comments. The authors would also like to thank Philippe Niger, Curtis Sherbo, and Michael Busser for reviewing this document. Other thanks go to Bharat Joshi, Mark Townsley, Wojciech Dec, and Kim Hylgaard who have had valuable comments during the development of this work.
[ANCP-FRAME] Ooghe, S., Voigt, N., Platnic, M., Haag, T., and S. Wadhwa, "Framework and Requirements for an Access Node Control Mechanism in Broadband Multi-Service Networks", Work in Progress, October 2009.
[ANCP-MIB] De Cnodder, S. and M. Morgenstern, "Access Node Control Protocol (ANCP) MIB module for Access Nodes", Work in Progress, July 2009.
Moustafa, et al. Informational [Page 17]
RFC 5713 ANCP Threats January 2010
Authors' Addresses
Hassnaa Moustafa France Telecom 38-40 rue du General Leclerc Issy Les Moulineaux, 92794 Cedex 9 France
EMail: hassnaa.moustafa@orange-ftgroup.com
Hannes Tschofenig Nokia Siemens Networks Linnoitustie 6 Espoo 02600 Finland