Internet Engineering Task Force (IETF) T. Mizrahi Request for Comments: 7384 Marvell Category: Informational October 2014 ISSN: 2070-1721
Security Requirements of Time Protocols in Packet Switched Networks
Abstract
As time and frequency distribution protocols are becoming increasingly common and widely deployed, concern about their exposure to various security threats is increasing. This document defines a set of security requirements for time protocols, focusing on the Precision Time Protocol (PTP) and the Network Time Protocol (NTP). This document also discusses the security impacts of time protocol practices, the performance implications of external security practices on time protocols, and the dependencies between other security services and time synchronization.
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/rfc7384.
Mizrahi Informational [Page 1]
RFC 7384 Time Protocol Security Requirements October 2014
Copyright Notice
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.
Table of Contents
1. Introduction ....................................................4 2. Terminology .....................................................5 2.1. Requirements Language ......................................5 2.2. Abbreviations ..............................................6 2.3. Common Terminology for PTP and NTP .........................6 2.4. Terms Used in This Document ................................6 3. Security Threats ................................................7 3.1. Threat Model ...............................................8 3.1.1. Internal vs. External Attackers .....................8 3.1.2. Man in the Middle (MITM) vs. Packet Injector ........8 3.2. Threat Analysis ............................................9 3.2.1. Packet Manipulation .................................9 3.2.2. Spoofing ............................................9 3.2.3. Replay Attack .......................................9 3.2.4. Rogue Master Attack .................................9 3.2.5. Packet Interception and Removal ....................10 3.2.6. Packet Delay Manipulation ..........................10 3.2.7. L2/L3 DoS Attacks ..................................10 3.2.8. Cryptographic Performance Attacks ..................10 3.2.9. DoS Attacks against the Time Protocol ..............11 3.2.10. Grandmaster Time Source Attack (e.g., GPS Fraud) ..11 3.2.11. Exploiting Vulnerabilities in the Time Protocol ...11 3.2.12. Network Reconnaissance ............................11 3.3. Threat Analysis Summary ...................................12 4. Requirement Levels .............................................13 5. Security Requirements ..........................................14 5.1. Clock Identity Authentication and Authorization ...........14 5.1.1. Authentication and Authorization of Masters ........15 5.1.2. Recursive Authentication and Authorization of Masters (Chain of Trust) ........................16 5.1.3. Authentication and Authorization of Slaves .........17
Mizrahi Informational [Page 2]
RFC 7384 Time Protocol Security Requirements October 2014
5.1.4. PTP: Authentication and Authorization of P2P TCs by the Master ..............................18 5.1.5. PTP: Authentication and Authorization of Control Messages ...................................18 5.2. Protocol Packet Integrity .................................19 5.2.1. PTP: Hop-by-Hop vs. End-to-End Integrity Protection .........................................20 5.2.1.1. Hop-by-Hop Integrity Protection ...........20 5.2.1.2. End-to-End Integrity Protection ...........21 5.3. Spoofing Prevention .......................................21 5.4. Availability ..............................................22 5.5. Replay Protection .........................................23 5.6. Cryptographic Keys and Security Associations ..............23 5.6.1. Key Freshness ......................................23 5.6.2. Security Association ...............................24 5.6.3. Unicast and Multicast Associations .................24 5.7. Performance ...............................................25 5.8. Confidentiality ...........................................26 5.9. Protection against Packet Delay and Interception Attacks ..27 5.10. Combining Secured with Unsecured Nodes ...................27 5.10.1. Secure Mode .......................................28 5.10.2. Hybrid Mode .......................................28 6. Summary of Requirements ........................................29 7. Additional Security Implications ...............................31 7.1. Security and On-the-Fly Timestamping ......................31 7.2. PTP: Security and Two-Step Timestamping ...................31 7.3. Intermediate Clocks .......................................32 7.4. External Security Protocols and Time Protocols ............32 7.5. External Security Services Requiring Time .................33 7.5.1. Timestamped Certificates ...........................33 7.5.2. Time Changes and Replay Attacks ....................33 8. Issues for Further Discussion ..................................34 9. Security Considerations ........................................34 10. References ....................................................34 10.1. Normative References .....................................34 10.2. Informative References ...................................34 Acknowledgments ...................................................36 Contributors ......................................................36 Author's Address ..................................................36
Mizrahi Informational [Page 3]
RFC 7384 Time Protocol Security Requirements October 2014
As time protocols are becoming increasingly common and widely deployed, concern about the resulting exposure to various security threats is increasing. If a time protocol is compromised, the applications it serves are prone to a range of possible attacks including Denial of Service (DoS) or incorrect behavior.
This document discusses the security aspects of time distribution protocols in packet networks and focuses on the two most common protocols: the Network Time Protocol [NTPv4] and the Precision Time Protocol (PTP) [IEEE1588]. Note that although PTP was not defined by the IETF, it is one of the two most common time protocols; hence, it is included in the discussion.
The Network Time Protocol was defined with an inherent security protocol; [NTPv4] defines a security protocol that is based on a symmetric key authentication scheme, and [AutoKey] presents an alternative security protocol, based on a public key authentication scheme. [IEEE1588] includes an experimental security protocol, defined in Annex K of the standard, but this Annex was never formalized into a fully defined security protocol.
While NTP includes an inherent security protocol, the absence of a standard security solution for PTP undoubtedly contributed to the wide deployment of unsecured time synchronization solutions. However, in some cases, security mechanisms may not be strictly necessary, e.g., due to other security practices in place or due to the architecture of the network. A time synchronization security solution, much like any security solution, is comprised of various building blocks and must be carefully tailored for the specific system in which it is deployed. Based on a system-specific threat assessment, the benefits of a security solution must be weighed against the potential risks, and based on this trade-off an optimal security solution can be selected.
The target audience of this document includes:
o Timing and networking equipment vendors - can benefit from this document by deriving the security features that should be supported in the time/networking equipment.
o Standards development organizations - can use the requirements defined in this document when specifying security mechanisms for a time protocol.
Mizrahi Informational [Page 4]
RFC 7384 Time Protocol Security Requirements October 2014
o Network operators - can use this document as a reference when designing a network and its security architecture. As stated above, the requirements in this document may be deployed selectively based on a careful per-system threat analysis.
This document attempts to add clarity to the time protocol security requirements discussion by addressing a series of questions:
(1) What are the threats that need to be addressed for the time protocol and what security services need to be provided (e.g., a malicious NTP server or PTP master)?
(2) What external security practices impact the security and performance of time keeping and what can be done to mitigate these impacts (e.g., an IPsec tunnel in the time protocol traffic path)?
(3) What are the security impacts of time protocol practices (e.g., on-the-fly modification of timestamps)?
(4) What are the dependencies between other security services and time protocols? (For example, which comes first - the certificate or the timestamp?)
In light of the questions above, this document defines a set of requirements for security solutions for time protocols, focusing on PTP and NTP.
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 [KEYWORDS].
This document describes security requirements; thus, requirements are phrased in the document in the form "the security mechanism MUST/SHOULD/...". Note that the phrasing does not imply that this document defines a specific security mechanism, but that it defines the requirements with which every security mechanism should comply.
Mizrahi Informational [Page 5]
RFC 7384 Time Protocol Security Requirements October 2014
This document refers to both PTP and NTP. For the sake of consistency, throughout the document the term "master" applies to both a PTP master and an NTP server. Similarly, the term "slave" applies to both PTP slaves and NTP clients. The term "protocol packets" refers generically to PTP and NTP messages.
o Clock - A node participating in the protocol (either PTP or NTP). A clock can be a master, a slave, or an intermediate clock (see corresponding definitions below).
o Control packets - Packets used by the protocol to exchange information between clocks that is not strictly related to the time. NTP uses NTP Control Messages. PTP uses Announce, Signaling, and Management messages.
o End-to-end security - A security approach where secured packets sent from a source to a destination are not modified by intermediate nodes, allowing the destination to authenticate the source of the packets and to verify their integrity. In the context of confidentiality, end-to-end encryption guarantees that intermediate nodes cannot eavesdrop to en route packets. However, as discussed in Section 5, confidentiality is not a strict requirement in this document.
Mizrahi Informational [Page 6]
RFC 7384 Time Protocol Security Requirements October 2014
o Grandmaster - A master that receives time information from a locally attached clock device and not through the network. A grandmaster distributes its time to other clocks in the network.
o Hop-by-hop security - A security approach where secured packets sent from a source to a destination may be modified by intermediate nodes. In this approach intermediate nodes share the encryption key with the source and destination, allowing them to re-encrypt or re-authenticate modified packets before relaying them to the destination.
o Intermediate clock - A clock that receives timing information from a master and sends timing information to other clocks. In NTP, this term refers to an NTP server that is not a Stratum 1 server. In PTP, this term refers to a BC or a TC.
o Master - A clock that generates timing information to other clocks in the network. In NTP, 'master' refers to an NTP server. In PTP, 'master' refers to a master OC (aka grandmaster) or to a port of a BC that is in the master state.
o Protocol packets - Packets used by the time protocol. The terminology used in this document distinguishes between time packets and control packets.
o Secured clock - A clock that supports a security mechanism that complies to the requirements in this document.
o Slave - A clock that receives timing information from a master. In NTP, 'slave' refers to an NTP client. In PTP, 'slave' refers to a slave OC or to a port of a BC that is in the slave state.
o Time packets - Protocol packets carrying time information.
o Unsecured clock - A clock that does not support a security mechanism according to the requirements in this document.
This section discusses the possible attacker types and analyzes various attacks against time protocols.
The literature is rich with security threats of time protocols, e.g., [Traps], [AutoKey], [TimeSec], [SecPTP], and [SecSen]. The threat analysis in this document is mostly based on [TimeSec].
Mizrahi Informational [Page 7]
RFC 7384 Time Protocol Security Requirements October 2014
In the context of internal and external attackers, the underlying assumption is that the time protocol is secured by either an encryption mechanism, an authentication mechanism, or both.
Internal attackers either have access to a trusted segment of the network or possess the encryption or authentication keys. An internal attack can also be performed by exploiting vulnerabilities in devices; for example, by installing malware or obtaining credentials to reconfigure the device. Thus, an internal attacker can maliciously tamper with legitimate traffic in the network as well as generate its own traffic and make it appear legitimate to its attacked nodes.
Note that internal attacks are a special case of Byzantine failures, where a node in the system may fail in arbitrary ways; by crashing, by omitting messages, or by malicious behavior. This document focuses on nodes that demonstrate malicious behavior.
External attackers, on the other hand, do not have the keys and have access only to the encrypted or authenticated traffic.
Obviously, in the absence of a security mechanism, there is no distinction between internal and external attackers, since all attackers are internal in practice.
3.1.2. Man in the Middle (MITM) vs. Packet Injector
MITM attackers are located in a position that allows interception and modification of in-flight protocol packets. It is assumed that an MITM attacker has physical access to a segment of the network or has gained control of one of the nodes in the network.
A traffic injector is not located in an MITM position, but can attack by generating protocol packets. An injector can reside either within the attacked network or on an external network that is connected to the attacked network. An injector can also potentially eavesdrop on protocol packets sent as multicast, record them, and replay them later.
Mizrahi Informational [Page 8]
RFC 7384 Time Protocol Security Requirements October 2014
A packet manipulation attack results when an MITM attacker receives timing protocol packets, alters them, and relays them to their destination, allowing the attacker to maliciously tamper with the protocol. This can result in a situation where the time protocol is apparently operational but providing intentionally inaccurate information.
In spoofing, an injector masquerades as a legitimate node in the network by generating and transmitting protocol packets or control packets. Two typical examples of spoofing attacks:
o An attacker can impersonate the master, allowing malicious distribution of false timing information.
o An attacker can impersonate a legitimate clock, a slave, or an intermediate clock, by sending malicious messages to the master, causing the master to respond to the legitimate clock with protocol packets that are based on the spoofed messages. Consequently, the delay computations of the legitimate clock are based on false information.
As with packet manipulation, this attack can result in a situation where the time protocol is apparently operational but providing intentionally inaccurate information.
In a replay attack, an attacker records protocol packets and replays them at a later time without any modification. This can also result in a situation where the time protocol is apparently operational but providing intentionally inaccurate information.
In a rogue master attack, an attacker causes other nodes in the network to believe it is a legitimate master. As opposed to the spoofing attack, in the rogue master attack the attacker does not fake its identity, but rather manipulates the master election process using malicious control packets. For example, in PTP, an attacker can manipulate the Best Master Clock Algorithm (BMCA) and cause other nodes in the network to believe it is the most eligible candidate to be a grandmaster.
Mizrahi Informational [Page 9]
RFC 7384 Time Protocol Security Requirements October 2014
In PTP, a possible variant of this attack is the rogue TC/BC attack. Similar to the rogue master attack, an attacker can cause victims to believe it is a legitimate TC or BC, allowing the attacker to manipulate the time information forwarded to the victims.
A packet interception and removal attack results when an MITM attacker intercepts and drops protocol packets, preventing the destination node from receiving some or all of the protocol packets.
In a packet delay manipulation scenario, an MITM attacker receives protocol packets and relays them to their destination after adding a maliciously computed delay. The attacker can use various delay attack strategies; the added delay can be constant, jittered, or slowly wandering. Each of these strategies has a different impact, but they all effectively manipulate the attacked clock.
Note that the victim still receives one copy of each packet, contrary to the replay attack, where some or all of the packets may be received by the victim more than once.
There are many possible Layer 2 and Layer 3 DoS attacks, e.g., IP spoofing, ARP spoofing [Hack], MAC flooding [Anatomy], and many others. As the target's availability is compromised, the timing protocol is affected accordingly.
In cryptographic performance attacks, an attacker transmits fake protocol packets, causing high utilization of the cryptographic engine at the receiver, which attempts to verify the integrity of these fake packets.
This DoS attack is applicable to all encryption and authentication protocols. However, when the time protocol uses a dedicated security mechanism implemented in a dedicated cryptographic engine, this attack can be applied to cause DoS specifically to the time protocol.
Mizrahi Informational [Page 10]
RFC 7384 Time Protocol Security Requirements October 2014
An attacker can attack a clock by sending an excessive number of time protocol packets, thus degrading the victim's performance. This attack can be implemented, for example, using the attacks described in Sections 3.2.2 and 3.2.4.
3.2.10. Grandmaster Time Source Attack (e.g., GPS Fraud)
Grandmasters receive their time from an external accurate time source, such as an atomic clock or a GPS clock, and then distribute this time to the slaves using the time protocol.
Time source attacks are aimed at the accurate time source of the grandmaster. For example, if the grandmaster uses a GPS-based clock as its reference source, an attacker can jam the reception of the GPS signal, or transmit a signal similar to one from a GPS satellite, causing the grandmaster to use a false reference time.
Note that this attack is outside the scope of the time protocol. While various security measures can be taken to mitigate this attack, these measures are outside the scope of the security requirements defined in this document.
3.2.11. Exploiting Vulnerabilities in the Time Protocol
Time protocols can be attacked by exploiting vulnerabilities in the protocol, implementation bugs, or misconfigurations (e.g., [NTPDDoS]). It should be noted that such attacks cannot typically be mitigated by security mechanisms. However, when a new vulnerability is discovered, operators should react as soon as possible, and take the necessary measures to address it.
An attacker can exploit the time protocol to collect information such as addresses and locations of nodes that take part in the protocol. Reconnaissance can be applied by either passively eavesdropping on protocol packets or sending malicious packets and gathering information from the responses. By eavesdropping on a time protocol, an attacker can learn the network latencies, which provide information about the network topology and node locations.
Moreover, properties such as the frequency of the protocol packets, or the exact times at which they are sent, can allow fingerprinting of specific nodes; thus, protocol packets from a node can be identified even if network addresses are hidden or encrypted.
Mizrahi Informational [Page 11]
RFC 7384 Time Protocol Security Requirements October 2014
The two key factors to a threat analysis are the impact and the likelihood of each of the analyzed attacks.
Table 1 summarizes the security attacks presented in Section 3.2. For each attack, the table specifies its impact, and its applicability to each of the attacker types presented in Section 3.1.
Table 1 clearly shows the distinction between external and internal attackers, and motivates the usage of authentication and integrity protection, significantly reducing the impact of external attackers.
The Impact column provides an intuitive measure of the severity of each attack, and the relevant Attacker Type column provides an intuition about how difficult each attack is to implement and, hence, about the likelihood of each attack.
The Impact column in Table 1 can have one of three values:
o DoS - the attack causes denial of service to the attacked node, the impact of which is not restricted to the time protocol.
o Accuracy degradation - the attack yields a degradation in the slave accuracy, but does not completely compromise the slaves' time and frequency.
o False time - slaves align to a false time or frequency value due to the attack. Note that if the time protocol aligns to a false time, it may cause DoS to other applications that rely on accurate time. However, for the purpose of the analysis in this section, we distinguish this implication from 'DoS', which refers to a DoS attack that is not necessarily aimed at the time protocol. All attacks that have a '+' for 'False Time' implicitly have a '+' for 'Accuracy Degradation'. Note that 'False Time' necessarily implies 'Accuracy Degradation'. However, two different terms are used, indicating two levels of severity.
The Attacker Type column refers to the four possible combinations of the attacker types defined in Section 3.1.
Mizrahi Informational [Page 12]
RFC 7384 Time Protocol Security Requirements October 2014
The security requirements are presented in Section 5. Each requirement is defined with a requirement level, in accordance with the requirement levels defined in Section 2.1.
The requirement levels in this document are affected by the following factors:
o Impact: The possible impact of not implementing the requirement, as illustrated in the Impact column of Table 1. For example, a requirement that addresses a threat that can be implemented by an external injector is typically a 'MUST', since the threat can be implemented by all the attacker types analyzed in Section 3.1.
Mizrahi Informational [Page 13]
RFC 7384 Time Protocol Security Requirements October 2014
o Difficulty of the corresponding attack: The level of difficulty of the possible attacks that become possible by not implementing the requirement. The level of difficulty is reflected in the Attacker Type column of Table 1. For example, a requirement that addresses a threat that only compromises the availability of the protocol is typically no more than a 'SHOULD'.
o Practical considerations: Various practical factors that may affect the requirement. For example, if a requirement is very difficult to implement, or is applicable to very specific scenarios, these factors may reduce the requirement level.
Section 5 lists the requirements. For each requirement, there is a short explanation detailing the reason for its requirement level.
This section defines a set of security requirements. These requirements are phrased in the form "the security mechanism MUST/SHOULD/MAY...". However, this document does not specify how these requirements can be met. While these requirements can be satisfied by defining explicit security mechanisms for time protocols, at least a subset of the requirements can be met by applying common security practices to the network or by using existing security protocols, such as [IPsec] or [MACsec]. Thus, security solutions that address these requirements are outside the scope of this document.
5.1. Clock Identity Authentication and Authorization
Requirement
The security mechanism MUST support authentication.
Requirement
The security mechanism MUST support authorization.
Requirement Level
The requirements in this subsection address the spoofing attack (Section 3.2.2) and the rogue master attack (Section 3.2.4).
The requirement level of these requirements is 'MUST' since, in the absence of these requirements, the protocol is exposed to attacks that are easy to implement and have a high impact.
Mizrahi Informational [Page 14]
RFC 7384 Time Protocol Security Requirements October 2014
Discussion
Authentication refers to verifying the identity of the peer clock. Authorization, on the other hand, refers to verifying that the peer clock is permitted to play the role that it plays in the protocol. For example, some nodes may be permitted to be masters, while other nodes are only permitted to be slaves or TCs.
Authentication is typically implemented by means of a cryptographic signature, allowing the verification of the identity of the sender. Authorization requires clocks to maintain a list of authorized clocks, or a "black list" of clocks that should be denied service or revoked.
It is noted that while the security mechanism is required to provide an authorization mechanism, the deployment of such a mechanism depends on the nature of the network. For example, a network that deploys PTP may consist of a set of identical OCs, where all clocks are equally permitted to be a master. In such a network, an authorization mechanism may not be necessary.
The following subsections describe five distinct cases of clock authentication.
5.1.1. Authentication and Authorization of Masters
Requirement
The security mechanism MUST support an authentication mechanism, allowing slaves to authenticate the identity of masters.
Requirement
The authentication mechanism MUST allow slaves to verify that the authenticated master is authorized to be a master.
Requirement Level
The requirements in this subsection address the spoofing attack (Section 3.2.2) and the rogue master attack (Section 3.2.4).
The requirement level of these requirements is 'MUST' since, in the absence of these requirements, the protocol is exposed to attacks that are easy to implement and have a high impact.
Mizrahi Informational [Page 15]
RFC 7384 Time Protocol Security Requirements October 2014
Discussion
Clocks authenticate masters in order to ensure the authenticity of the time source. It is important for a slave to verify the identity of the master, as well as to verify that the master is indeed authorized to be a master.
5.1.2. Recursive Authentication and Authorization of Masters (Chain of Trust)
Requirement
The security mechanism MUST support recursive authentication and authorization of the master, to be used in cases where time information is conveyed through intermediate clocks.
Requirement Level
The requirement in this subsection addresses the spoofing attack (Section 3.2.2) and the rogue master attack (Section 3.2.4).
The requirement level of this requirement is 'MUST' since, in the absence of this requirement, the protocol is exposed to attacks that are easy to implement and have a high impact.
Discussion
In some cases, a slave is connected to an intermediate clock that is not the primary time source. For example, in PTP, a slave can be connected to a Boundary Clock (BC) or a Transparent Clock (TC), which in turn is connected to a grandmaster. A similar example in NTP is when a client is connected to a Stratum 2 server, which is connected to a Stratum 1 server. In both the PTP and the NTP cases, the slave authenticates the intermediate clock, and the intermediate clock authenticates the grandmaster. This recursive authentication process is referred to in [AutoKey] as proventication.
Specifically in PTP, this requirement implies that if a slave receives time information through a TC, it must authenticate the TC to which it is attached, as well as authenticate the master from which it receives the time information, as per Section 5.1.1. Similarly, if a TC receives time information through an attached TC, it must authenticate the attached TC.
Mizrahi Informational [Page 16]
RFC 7384 Time Protocol Security Requirements October 2014
The security mechanism MAY provide a means for a master to authenticate its slaves.
Requirement
The security mechanism MAY provide a means for a master to verify that the sender of a protocol packet is authorized to send a packet of this type.
Requirement Level
The requirement in this subsection prevents DoS attacks against the master (Section 3.2.9).
The requirement level of this requirement is 'MAY' since:
o Its impact is low, i.e., in the absence of this requirement the protocol is only exposed to DoS.
o Practical considerations: requiring an NTP server to authenticate its clients may significantly impose on the server's performance.
Note that while the requirement level of this requirement is 'MAY', the requirement in Section 5.1.1 is 'MUST'; the security mechanism must provide a means for authentication and authorization, with an emphasis on the master. Authentication and authorization of slaves are specified in this subsection as 'MAY'.
Discussion
Slaves and intermediate clocks are authenticated by masters in order to verify that they are authorized to receive timing services from the master.
Authentication of slaves prevents unauthorized clocks from receiving time services. Preventing the master from serving unauthorized clocks can help in mitigating DoS attacks against the master. Note that the authentication of slaves might put a higher load on the master than serving the unauthorized clock; hence, this requirement is 'MAY'.
Mizrahi Informational [Page 17]
RFC 7384 Time Protocol Security Requirements October 2014
5.1.4. PTP: Authentication and Authorization of P2P TCs by the Master
Requirement
The security mechanism for PTP MAY provide a means for a master to authenticate the identity of the P2P TCs directly connected to it.
Requirement
The security mechanism for PTP MAY provide a means for a master to verify that P2P TCs directly connected to it are authorized to be TCs.
Requirement Level
The requirement in this subsection prevents DoS attacks against the master (Section 3.2.9).
The requirement level of this requirement is 'MAY' for the same reasons specified in Section 5.1.3.
Discussion
P2P TCs that are one hop from the master use the PDelay_Req and PDelay_Resp handshake to compute the link delay between the master and TC. These TCs are authenticated by the master.
Authentication of TCs, much like authentication of slaves, reduces unnecessary load on the master and peer TCs, by preventing the master from serving unauthorized clocks.
5.1.5. PTP: Authentication and Authorization of Control Messages
Requirement
The security mechanism for PTP MUST support authentication of Announce messages. The authentication mechanism MUST also verify that the sender is authorized to be a master.
Requirement
The security mechanism for PTP MUST support authentication and authorization of Management messages.
Requirement
The security mechanism MAY support authentication and authorization of Signaling messages.
Mizrahi Informational [Page 18]
RFC 7384 Time Protocol Security Requirements October 2014
Requirement Level
The requirements in this subsection address the spoofing attack (Section 3.2.2) and the rogue master attack (Section 3.2.4).
The requirement level of the first two requirements is 'MUST' since, in the absence of these requirements, the protocol is exposed to attacks that are easy to implement and have a high impact.
The requirement level of the third requirement is 'MAY' since its impact greatly depends on the application for which the Signaling messages are used.
Discussion
Master election is performed in PTP using the Best Master Clock Algorithm (BMCA). Each Ordinary Clock (OC) announces its clock attributes using Announce messages, and the best master is elected based on the information gathered from all the candidates. Announce messages must be authenticated in order to prevent rogue master attacks (Section 3.2.4). Note that this subsection specifies a requirement that is not necessarily included in Sections 5.1.1 or 5.1.3, since the BMCA is initiated before clocks have been defined as masters or slaves.
Management messages are used to monitor or configure PTP clocks. Malicious usage of Management messages enables various attacks, such as the rogue master attack or DoS attack.
Signaling messages are used by PTP clocks to exchange information that is not strictly related to time information or to master selection, such as unicast negotiation. Authentication and authorization of Signaling messages may be required in some systems, depending on the application for which these messages are used.
The security mechanism MUST protect the integrity of protocol packets.
Mizrahi Informational [Page 19]
RFC 7384 Time Protocol Security Requirements October 2014
Requirement Level
The requirement in this subsection addresses the packet manipulation attack (Section 3.2.1).
The requirement level of this requirement is 'MUST' since, in the absence of this requirement, the protocol is exposed to attacks that are easy to implement and have high impact.
Discussion
While Section 5.1 refers to ensuring the identity an authorization of the source of a protocol packet, this subsection refers to ensuring that the packet arrived intact. The integrity protection mechanism ensures the authenticity and completeness of data from the data originator.
Integrity protection is typically implemented by means of an Integrity Check Value (ICV) that is included in protocol packets and is verified by the receiver.
5.2.1. PTP: Hop-by-Hop vs. End-to-End Integrity Protection
Specifically in PTP, when protocol packets are subject to modification by TCs, the integrity protection can be enforced in one of two approaches: end-to-end or hop-by-hop.
o Modifies the packet, i.e., modifies the correctionField. Note: TCs improve the end-to-end accuracy by updating a correctionField (Clause 6.5 in [IEEE1588]) in the PTP packet by adding the latency caused by the current TC.
o Re-generates the integrity protection, e.g., re-computes a Message Authentication Code (MAC).
In the hop-by-hop approach, the integrity of protocol packets is protected by induction on the path from the originator to the receiver.
This approach is simple, but allows rogue TCs to modify protocol packets.
Mizrahi Informational [Page 20]
RFC 7384 Time Protocol Security Requirements October 2014
In this approach, the integrity protection is maintained on the path from the originator of a protocol packet to the receiver. This allows the receiver to directly validate the protocol packet without the ability of intermediate TCs to manipulate the packet.
Since TCs need to modify the correctionField, a separate integrity protection mechanism is used specifically for the correctionField.
The end-to-end approach limits the TC's impact to the correctionField alone, while the rest of the protocol packet is protected on an end- to-end basis. It should be noted that this approach is more difficult to implement than the hop-by-hop approach, as it requires the correctionField to be protected separately from the other fields of the packet, possibly using different cryptographic mechanisms and keys.
The security mechanism MUST provide a means to prevent master spoofing.
Requirement
The security mechanism MUST provide a means to prevent slave spoofing.
Requirement
PTP: The security mechanism MUST provide a means to prevent P2P TC spoofing.
Requirement Level
The requirements in this subsection address spoofing attacks. As described in Section 3.2.2, when these requirements are not met, the attack may have a high impact, causing slaves to rely on false time information. Thus, the requirement level is 'MUST'.
Discussion
Spoofing attacks may take various forms, and they can potentially cause significant impact. In a master spoofing attack, the attacker causes slaves to receive false information about the current time by masquerading as the master.
Mizrahi Informational [Page 21]
RFC 7384 Time Protocol Security Requirements October 2014
By spoofing a slave or an intermediate node (the second example of Section 3.2.2), an attacker can tamper with the slaves' delay computations. These attacks can be mitigated by an authentication mechanism (Sections 5.1.3 and 5.1.4) or by other means, for example, a PTP Delay_Req can include a MAC that is included in the corresponding Delay_Resp message, allowing the slave to verify that the Delay_Resp was not sent in response to a spoofed message.
The security mechanism SHOULD include measures to mitigate DoS attacks against the time protocol.
Requirement Level
The requirement in this subsection prevents DoS attacks against the protocol (Section 3.2.9).
The requirement level of this requirement is 'SHOULD' due to its low impact, i.e., in the absence of this requirement the protocol is only exposed to DoS.
Discussion
The protocol availability can be compromised by several different attacks. An attacker can inject protocol packets to implement the spoofing attack (Section 3.2.2) or the rogue master attack (Section 3.2.4), causing DoS to the victim (Section 3.2.9).
An authentication mechanism (Section 5.1) limits these attacks strictly to internal attackers; thus, it prevents external attackers from performing them. Hence, the requirements of Section 5.1 can be used to mitigate this attack. Note that Section 5.1 addresses a wider range of threats, whereas the current section is focused on availability.
The DoS attacks described in Section 3.2.7 are performed at lower layers than the time protocol layer, and they are thus outside the scope of the security requirements defined in this document.
Mizrahi Informational [Page 22]
RFC 7384 Time Protocol Security Requirements October 2014
The security mechanism MUST include a replay prevention mechanism.
Requirement Level
The requirement in this subsection prevents replay attacks (Section 3.2.3).
The requirement level of this requirement is 'MUST' since, in the absence of this requirement, the protocol is exposed to attacks that are easy to implement and have a high impact.
Discussion
The replay attack (Section 3.2.3) can compromise both the integrity and availability of the protocol. Common encryption and authentication mechanisms include replay prevention mechanisms that typically use a monotonously increasing packet sequence number.
The security mechanism MUST provide a means to refresh the cryptographic keys.
The cryptographic keys MUST be refreshed frequently.
Requirement Level
The requirement level of this requirement is 'MUST' since key freshness is an essential property for cryptographic algorithms, as discussed below.
Discussion
Key freshness guarantees that both sides share a common updated secret key. It also helps in preventing replay attacks. Thus, it is important for keys to be refreshed frequently. Note that the term 'frequently' is used without a quantitative requirement, as the precise frequency requirement should be considered on a per- system basis, based on the threats and system requirements.
Mizrahi Informational [Page 23]
RFC 7384 Time Protocol Security Requirements October 2014
The security protocol SHOULD support a security association protocol where:
o Two or more clocks authenticate each other.
o The clocks generate and agree on a cryptographic session key.
Requirement
Each instance of the association protocol SHOULD produce a different session key.
Requirement Level
The requirement level of this requirement is 'SHOULD' since it may be expensive in terms of performance, especially in low-cost clocks.
Discussion
The security requirements in Sections 5.1 and 5.2 require usage of cryptographic mechanisms, deploying cryptographic keys. A security association (e.g., [IPsec]) is an important building block in these mechanisms.
It should be noted that in some cases, different security association mechanisms may be used at different levels of clock hierarchies. For example, the association between a Stratum 2 clock and a Stratum 3 clock in NTP may have different characteristics than an association between two clocks at the same stratum level. On a related note, in some cases, a hybrid solution may be used, where a subset of the network is not secured at all (see Section 5.10.2).
The security mechanism SHOULD support security association protocols for unicast and for multicast associations.
Mizrahi Informational [Page 24]
RFC 7384 Time Protocol Security Requirements October 2014
Requirement Level
The requirement level of this requirement is 'SHOULD' since it may be expensive in terms of performance, especially for low-cost clocks.
Discussion
A unicast protocol requires an association protocol between two clocks, whereas a multicast protocol requires an association protocol among two or more clocks, where one of the clocks is a master.
The security mechanism MUST be designed in such a way that it does not significantly degrade the quality of the time transfer.
Requirement
The mechanism SHOULD minimize computational load.
Requirement
The mechanism SHOULD minimize storage requirements of client state in the master.
Requirement
The mechanism SHOULD minimize the bandwidth overhead required by the security protocol.
Requirement Level
While the quality of the time transfer is clearly a 'MUST', the other three performance requirements are 'SHOULD', since some systems may be more sensitive to resource consumption than others; hence, these requirements should be considered on a per-system basis.
Discussion
Performance efficiency is important since client restrictions often dictate a low processing and memory footprint and because the server may have extensive fan-out.
Mizrahi Informational [Page 25]
RFC 7384 Time Protocol Security Requirements October 2014
Note that the performance requirements refer to a time-protocol- specific security mechanism. In systems where a security protocol is used for other types of traffic as well, this document does not place any performance requirements on the security protocol performance. For example, if IPsec encryption is used for securing all information between the master and slave node, including information that is not part of the time protocol, the requirements in this subsection are not necessarily applicable.
The security mechanism MAY provide confidentiality protection of the protocol packets.
Requirement Level
The requirement level of this requirement is 'MAY' since the absence of this requirement does not expose the protocol to severe threats, as discussed below.
Discussion
In the context of time protocols, confidentiality is typically of low importance, since timing information is usually not considered secret information.
Confidentiality can play an important role when service providers charge their customers for time synchronization services; thus, an encryption mechanism can prevent eavesdroppers from obtaining the service without payment. Note that these cases are, for now, rather esoteric.
Confidentiality can also prevent an MITM attacker from identifying protocol packets. Thus, confidentiality can assist in protecting the timing protocol against MITM attacks such as packet delay (Section 3.2.6), manipulation and interception, and removal attacks. Note that time protocols have predictable behavior even after encryption, such as packet transmission rates and packet lengths. Additional measures can be taken to mitigate encrypted traffic analysis by random padding of encrypted packets and by adding random dummy packets. Nevertheless, encryption does not prevent such MITM attacks, but rather makes these attacks more difficult to implement.
Mizrahi Informational [Page 26]
RFC 7384 Time Protocol Security Requirements October 2014
5.9. Protection against Packet Delay and Interception Attacks
Requirement
The security mechanism MUST include means to protect the protocol from MITM attacks that degrade the clock accuracy.
Requirement Level
The requirements in this subsection address MITM attacks such as the packet delay attack (Section 3.2.6) and packet interception attacks (Sections 3.2.5 and 3.2.1).
The requirement level of this requirement is 'MUST'. In the absence of this requirement, the protocol is exposed to attacks that are easy to implement and have a high impact. Note that in the absence of this requirement, the impact is similar to packet manipulation attacks (Section 3.2.1); thus, this requirement has the same requirement level as integrity protection (Section 5.2).
It is noted that the implementation of this requirement depends on the topology and properties of the system.
Discussion
While this document does not define specific security solutions, we note that common practices for protection against MITM attacks use redundant masters (e.g., [NTPv4]) or redundant paths between the master and slave (e.g., [DelayAtt]). If one of the time sources indicates a time value that is significantly different than the other sources, it is assumed to be erroneous or under attack and is therefore ignored.
Thus, MITM attack prevention derives a requirement from the security mechanism and a requirement from the network topology. While the security mechanism should support the ability to detect delay attacks, it is noted that in some networks it is not possible to provide the redundancy needed for such a detection mechanism.
Integrating a security mechanism into a time-synchronized system is a complex and expensive process, and hence in some cases may require incremental deployment, where new equipment supports the security mechanism, and is required to interoperate with legacy equipment without the security features.
Mizrahi Informational [Page 27]
RFC 7384 Time Protocol Security Requirements October 2014
The security mechanism MUST support a secure mode, where only secured clocks are permitted to take part in the time protocol. In this mode every protocol packet received from an unsecured clock MUST be discarded.
Requirement Level
The requirement level of this requirement is 'MUST' since the full capacity of the security requirements defined in this document can only be achieved in secure mode.
Discussion
While the requirement in this subsection is similar to the one in Section 5.1, it refers to the secure mode, as opposed to the hybrid mode presented in the next subsection.
The security protocol SHOULD support a hybrid mode, where both secured and unsecured clocks are permitted to take part in the protocol.
Requirement Level
The requirement level of this requirement is 'SHOULD'; on one hand, hybrid mode enables a gradual transition from unsecured to secured mode, which is especially important in large-scaled deployments. On the other hand, hybrid mode is not required in all systems; this document recommends deployment of the 'secure mode' described in Section 5.10.1, where possible.
Discussion
The hybrid mode allows both secured and unsecured clocks to take part in the time protocol. NTP, for example, allows a mixture of secured and unsecured nodes.
Requirement
A master in the hybrid mode SHOULD be a secured clock.
Mizrahi Informational [Page 28]
RFC 7384 Time Protocol Security Requirements October 2014
A secured slave in the hybrid mode SHOULD discard all protocol packets received from unsecured clocks.
Requirement Level
The requirement level of this requirement is 'SHOULD' since it may not be applicable to all deployments. For example, a hybrid network may require the usage of unsecured masters or TCs.
Discussion
This requirement ensures that the existence of unsecured clocks does not compromise the security provided to secured clocks. Hence, secured slaves only "trust" protocol packets received from a secured clock.
An unsecured slave can receive protocol packets from either unsecured clocks or secured clocks. Note that the latter does not apply when encryption is used. When integrity protection is used, the unsecured slave can receive secured packets ignoring the integrity protection.
Note that the security scheme in [NTPv4] with [AutoKey] does not satisfy this requirement, since nodes prefer the server with the most accurate clock, which is not necessarily the server that supports authentication. For example, a Stratum 2 server is connected to two Stratum 1 servers: Server A, supporting authentication, and Server B, without authentication. If Server B has a more accurate clock than A, the Stratum 2 server chooses Server B, in spite of the fact it does not support authentication.
This section discusses additional implications of the interaction between time protocols and security mechanisms.
This section refers to time protocol security mechanisms, as well as to "external" security mechanisms, i.e., security mechanisms that are not strictly related to the time protocol.
Time protocols often require that protocol packets be modified during transmission. Both NTP and PTP in one-step mode require clocks to modify protocol packets based on the time of transmission and/or reception.
In the presence of a security mechanism, whether encryption or integrity protection:
o During transmission the encryption and/or integrity protection MUST be applied after integrating the timestamp into the packet.
To allow high accuracy, timestamping is typically performed as close to the transmission or reception time as possible. However, since the security engine must be placed between the timestamping function and the physical interface, it may introduce non-deterministic latency that causes accuracy degradation. These performance aspects have been analyzed in literature, e.g., [1588IPsec] and [Tunnel].
PTP supports a two-step mode of operation, where the time of transmission of protocol packets is communicated without modifying the packets. As opposed to one-step mode, two-step timestamping can be performed without the requirement to encrypt after timestamping.
Note that if an encryption mechanism such as IPsec is used, it presents a challenge to the timestamping mechanism, since time protocol packets are encrypted when traversing the physical interface, and are thus impossible to identify. A possible solution to this problem [IPsecSync] is to include an indication in the encryption header that identifies time protocol packets.
Mizrahi Informational [Page 31]
RFC 7384 Time Protocol Security Requirements October 2014
A time protocol allows slaves to receive time information from an accurate time source. Time information is sent over a path that often traverses one or more intermediate clocks.
o In NTP, time information originated from a Stratum 1 server can be distributed to Stratum 2 servers and, in turn, distributed from the Stratum 2 servers to NTP clients. In this case, the Stratum 2 servers are a layer of intermediate clocks. These intermediate clocks are referred to as "secondary servers" in [NTPv4].
o In PTP, BCs and TCs are intermediate nodes used to improve the accuracy of time information conveyed between the grandmaster and the slaves.
A common rule of thumb in network security is that end-to-end security is the best policy, as it secures the entire path between the data originator and its receiver. The usage of intermediate nodes implies that if a security mechanism is deployed in the network, a hop-by-hop security scheme must be used, since intermediate nodes must be able to send time information to the slaves, or to modify time information sent through them.
This inherent property of using intermediate clocks increases the system's exposure to internal threats, as a large number of nodes possess the security keys.
Thus, there is a trade-off between the achievable clock accuracy of a system, and the robustness of its security solution. On one hand, high clock accuracy calls for hop-by-hop involvement in the protocol, also known as on-path support. On the other hand, a robust security solution calls for end-to-end data protection.
7.4. External Security Protocols and Time Protocols
Time protocols are often deployed in systems that use security mechanisms and protocols.
A typical example is the 3GPP Femtocell network [3GPP], where IPsec is used for securing traffic between a Femtocell and the Femto Gateway. In some cases, all traffic between these two nodes may be secured by IPsec, including the time protocol traffic. This use-case is thoroughly discussed in [IPsecSync].
Another typical example is the usage of MACsec encryption ([MACsec]) in L2 networks that deploy time synchronization [AvbAssum].
Mizrahi Informational [Page 32]
RFC 7384 Time Protocol Security Requirements October 2014
The usage of external security mechanisms may affect time protocols as follows:
o Timestamping accuracy can be affected, as described in Section 7.1.
o If traffic is secured between two nodes in the network, no intermediate clocks can be used between these two nodes. In the [3GPP] example, if traffic between the Femtocell and the Femto Gateway is encrypted, then time protocol packets are necessarily transported over the underlying network without modification and, thus, cannot enjoy the improved accuracy provided by intermediate clock nodes.
Cryptographic protocols often use time as an important factor in the cryptographic algorithm. If a time protocol is compromised, it may consequently expose the security protocols that rely on it to various attacks. Two examples are presented in this section.
Certificate validation requires the sender and receiver to be roughly time synchronized. Thus, synchronization is required for establishing security protocols such as Internet Key Exchange Protocol version 2 (IKEv2) and Transport Layer Security (TLS). Other authentication and key exchange mechanisms, such as Kerberos, also require the parties involved to be synchronized [Kerb].
An even stronger interdependence between a time protocol and a security mechanism is defined in [AutoKey], which defines mutual dependence between the acquired time information, and the authentication protocol that secures it. This bootstrapping behavior results from the fact that trusting the received time information requires a valid certificate, and validating a certificate requires knowledge of the time.
A successful attack on a time protocol may cause the attacked clocks to go back in time. The erroneous time may expose cryptographic algorithms that rely on time, as a node may use a key that was already used in the past and has expired.
Mizrahi Informational [Page 33]
RFC 7384 Time Protocol Security Requirements October 2014
The Key distribution is outside the scope of this document. Although this is an essential element of any security system, it is outside the scope of this document.
[IEEE1588] IEEE, "1588-2008 - IEEE Standard for a Precision Clock Synchronization Protocol for Networked Measurement and Control Systems", IEEE Standard 1588-2008, July 2008.
[NTPv4] Mills, D., Martin, J., Ed., Burbank, J., and W. Kasch, "Network Time Protocol Version 4: Protocol and Algorithms Specification", RFC 5905, June 2010, <http://www.rfc-editor.org/info/rfc5905>.
[1588IPsec] Treytl, A. and B. Hirschler, "Securing IEEE 1588 by IPsec tunnels - An analysis", in Proceedings of 2010 International Symposium for Precision Clock Synchronization for Measurement, Control and Communication, ISPCS 2010, pp. 83-90, September 2010.
[3GPP] 3GPP, "Security of Home Node B (HNB) / Home evolved Node B (HeNB)", 3GPP TS 33.320 11.6.0, November 2012.
[Anatomy] Nachreiner, C., "Anatomy of an ARP Poisoning Attack", 2003.
RFC 7384 Time Protocol Security Requirements October 2014
[AvbAssum] Pannell, D., "Audio Video Bridging Gen 2 Assumptions", IEEE 802.1 AVB Plenary, Work in Progress, May 2012.
[DelayAtt] Mizrahi, T., "A game theoretic analysis of delay attacks against time synchronization protocols", accepted, to appear in Proceedings of the International IEEE Symposium on Precision Clock Synchronization for Measurement, Control and Communication, ISPCS, September 2012.
[Hack] McClure, S., Scambray, J., and G. Kurtz, "Hacking Exposed: Network Security Secrets and Solutions", McGraw-Hill, 2009.
[IPsecSync] Xu, Y., "IPsec security for packet based synchronization", Work in Progress, draft-xu-tictoc- ipsec-security-for-synchronization-02, September 2011.
[MACsec] IEEE, "IEEE Standard for Local and metropolitan area networks - Media Access Control (MAC) Security", IEEE Standard 802.1AE, August 2006.
[NTPDDoS] "Attackers use NTP reflection in huge DDoS attack", TICTOC mail archive, 2014.
[SecPTP] Tsang, J. and K. Beznosov, "A Security Analysis of the Precise Time Protocol (Short Paper)," 8th International Conference on Information and Communication Security (ICICS) Lecture Notes in Computer Science Volume 4307, pp. 50-59, 2006.
[SecSen] Ganeriwal, S., Popper, C., Capkun, S., and M. B. Srivastava, "Secure Time Synchronization in Sensor Networks", ACM Trans. Inf. Syst. Secur., Volume 11, Issue 4, Article 23, July 2008.
[TimeSec] Mizrahi, T., "Time synchronization security using IPsec and MACsec", ISPCS 2011, pp. 38-43, September 2011.
Mizrahi Informational [Page 35]
RFC 7384 Time Protocol Security Requirements October 2014
[Traps] Treytl, A., Gaderer, G., Hirschler, B., and R. Cohen, "Traps and pitfalls in secure clock synchronization" in Proceedings of 2007 International Symposium for Precision Clock Synchronization for Measurement, Control and Communication, ISPCS 2007, pp. 18-24, October 2007.
[Tunnel] Treytl, A., Hirschler, B., and T. Sauter, "Secure tunneling of high-precision clock synchronisation protocols and other time-stamped data", in Proceedings of the 8th IEEE International Workshop on Factory Communication Systems (WFCS), pp. 303-313, May 2010.
Acknowledgments
The author gratefully acknowledges Stefano Ruffini, Doug Arnold, Kevin Gross, Dieter Sibold, Dan Grossman, Laurent Montini, Russell Smiley, Shawn Emery, Dan Romascanu, Stephen Farrell, Kathleen Moriarty, and Joel Jaeggli for their thorough review and helpful comments. The author would also like to thank members of the TICTOC WG for providing feedback on the TICTOC mailing list.
Contributors
Karen O'Donoghue ISOC
EMail: odonoghue@isoc.org
Author's Address
Tal Mizrahi Marvell 6 Hamada St. Yokneam, 20692 Israel