Internet-Draft | General SAV Capabilities | August 2024 |
Huang, et al. | Expires 27 February 2025 | [Page] |
The SAV rules of existing source address validation (SAV) mechanisms, are derived from other core data structures, e.g., FIB-based uRPF, which are not dedicatedly designed for source filtering. Therefore there are some limitations related to deployable scenarios and traffic handling policies.¶
To overcome these limitations, this document introduces the general SAV capabilities from data plane perspective. How to implement the capabilities and how to generate SAV rules are not in the scope of this document.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 27 February 2025.¶
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Source address validation (SAV) can detect and prevent source address spoofing on the SAV-enabled routers. When a packet arrives at an interface of the router, the source address of the packet will be validated. The packets with unwanted source addresses or arriving at unwanted interfaces, will be considered invalid and usually be conducted elimination actions on. Only validated packets will continue to be handled or forwarded.¶
From the perspective of data plane validation, the SAV capabilities of existing mechanisms have two main limitations. One of them is the deployable scenario limitation. ACL rules can be configured for filtering unwanted source addresses at specific interfaces ([RFC3704]). However, ACL is not dedicatedly designed for source prefix filtering. There exist performance and scalability issues due to long-key based searching, and usually expert maintenance efforts are required. Strict uRPF and loose uRPF are two typical FIB-based SAV mechanisms ([RFC3704]) and are supported by most commercial routers/switches. FIB-based validation brings many benefits compared to ACL-based filtering but also induces some limitations. Strict uRPF is not applicable for asymmetric routing ([RFC8704]), which exists in various scenarios such as intra-domain multi-homing access, inter-domain interconnection, etc. Under asymmetric routing, a source prefix may have a different incoming interface from the next-hop interface of the matched entry, or the source prefix does not exist in the FIB at all. Loose mode can only block unannounced prefix, which results in massive false negatives. Overall, existing ACL-based or FIB-based SAVs can only be applied to specific scenarios and cannot be adaptive to various scenarios (e.g., symmetric vs asymmetric).¶
The other limitation is inflexible traffic handling policy. The current common practice is just to silently drop the spoofed packets. We don’t know who benefits from this and who is the source. Further more, the clues of attacks are ignored, which could be very helpful for dealing with DDoS attacks etc.¶
The root cause of the above two limitations is that there is no tool specifically designed for source address filtering. That is, the capabilities of current tools are derived from other functions, e.g., FIB or ACL.¶
This document describes the general SAV capabilities that the data plane of SAV-enabled devices should have. Two kinds of capabilities will be introduced: validation mode and traffic handling policy. Validation modes describe how to apply validation in different scenarios. Traffic handling policies are the policies applied on the validated packets. By implementing the general SAV capabilities, the above two limitations of existing mechanisms can be overcome.¶
To achieve accurate and scalable source address validation, a dedicated SAV table for SAV rules is needed instead of using those derived from other functions, e.g., FIB or ACL. Note that the general SAV capabilities described in this document is decoupled with real implementation. Conforming implementations of this specification may differ, but the SAV outcomes SHOULD be equivalent to the described SAV capabilities. And also how to generate SAV rules is not the focus of this document.¶
Validation mode: The mode that describes the typical applications of SAV in a specific kind of scenarios. Different modes take effect in different scales and treat the validated packets differently.¶
SAV rule: The entry specifying the incoming interfaces of specific source addresses or source prefixes. The SAV rule expression might be slightly diffrent between validation modes.¶
Traffic handling policy: The policy taken on the packets validated by SAV.¶
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 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This section describes three validation modes (Mode1 in Section 2.1, Mode2 in Section 2.2, Mode3 in Section 2.3). These modes take effect in different scales and need corresponding SAV rules to validate spoofing packets. By choosing modes in different scenarios appropriately, the network can be protected as much as possible while not impacting the forwarding of legitimate packets.¶
Mode 1 is an interface-scale mode, which means it takes effect on a specific interface. The interface enabling Mode 1 is maintaining an interface-based prefix allowlist--SAV rule 1. Only the source prefixes recorded in the list will be considered valid, otherwise invalid.¶
Applying Mode 1 on an interface requires the complete knowledge of legitimate prefixes connected to the interface. Mode 1 is suitable to the closed-connected interfaces such as those connecting to a subnet, a stub AS, or a customer cone. Such a mode can efficiently prevent the connected network from spoofing source prefixes of other networks.¶
Strict uRPF based on FIB belongs to this mode. However, to overcome the limitation of asymmetric routing, native source prefix-based SAV rule expression is suggested. This is essential for new SAV mechanisms or architectures such as EFP-uRPF [RFC8704], BAR-SAV [I-D.ietf-sidrops-bar-sav], Intra-domain/Inter-domain SAVNET ([I-D.li-savnet-intra-domain-architecture], [I-D.wu-savnet-inter-domain-architecture]), etc.¶
In some cases, it may be difficult for an interface getting all the legitimate source prefixes. If any legitimate prefix is not included in the allowlist, packets with this source addresses arriving at the interface may be improperly blocked. For example, the interface with a default route or the interface connecting to the Internet through a provider AS can hardly promise to know all the legitimate source prefixes.¶
Mode 2 is also an interface-scale mode, which means it takes effect on a configured interface. An interface cannot enable Mode 1 and Mode 2 at the same time. The interface enabling Mode 2 is maintaining an interface-based prefix blocklist--SAV rule 2. The source prefixes recorded in the list will be considered invalid, otherwise valid.¶
This mode does not require the complete blocklist. If the packets with the specific source addresses need to be discarded, Mode 2 can be taken. Mode 2 is suitable for proactive filtering and reactive filtering. Usually the source prefixes that are sure to be invalid will be put into the blocklist, which is proactive filtering. Reactive filtering rules are usually installed in DDoS elimination for dropping packets with specific source addresses.¶
The prefix blocklist can be generated automatically, e.g., one of Intra-domain SAVNET architecture cases, blocking the incoming traffic with internal source prefixes. Or operators can configure the specific source prefixes to block from the interface. This is similar to ACL-based filtering, but more native SAV rule expression with better performance and scalability.¶
Mode 3 is a router-scale mode, which means it can validate traffic arriving at the router from all directions. The router enabling Mode 3 will record the protected source prefixes and maintain an interface list for each source prefix-SAV rule 3. The interface list of each source prefix may be an allowlist or a blocklist.¶
If a source prefix has an interface allowlist, the packet with this source prefix is considered valid only when its incoming interface is in the interface allowlist. Otherwise, the packet is considered invalid.¶
Conversely, If a source prefix has an interface blocklist, the packet with this source prefix is considered invalid only when its incoming interface is in the interface blocklist. Otherwise, the packet is considered valid. Either allowlist or blocklist here shares the same validating effects, the difference in reality mainly lies on how long the list would be. So generally we put these two in one mode, and thereafter illustrate the coresponding procedure only based on interface allowlist.¶
Mode 3 focuses on validating/protecting the interested source prefixes. Mode 3 is applicable to the scenario where multiple interfaces are availalbe to provide potential connection to specific source prefix. If some source prefixes are important, Mode 3 can also provide stricter and more efficient protection than Mode 1 and Mode 2 for these source prefixes.¶
Operators can configure the interface list for a specific source prefix, to prevent DDoS attack related to this source prefix. Or the interface list for specific prefixes can be generated automatically, e.g., one capability defined by Intra-domain and Inter-domain SAVNET architectures.¶
Mode 1 and Mode 2 are working on interface-level, while Mode 3 is for the router-level. Thus, there can be multiple modes configured on the same router. Mode 1 are most preferred if applicable (with best protection effect) and mutual exclusive with the other two modes, which means while an interface enabled Mode 1, the traffic for this interface don't need go through Mode 2 or Mode 3 -- while the validation result for Mode 2 is valid, the traffic still need go through Mode 3 if applicable. Figure 1 shows a comparison of different validation modes for dealing with source address validation.¶
+------------------------------------------------------------------------------------------+ + Mode | Scale | SAV rule | validation result + +------------------------------------------------------------------------------------------+ + 1 | interface | 1: interface-based source prefix allowlist | invalid if not matched + + 2 | interface | 2: interface-based source prefix blocklist | invalid if matched + + 3 | router | 3: prefix-based interface allowlist | invalid if not matched + +------------------------------------------------------------------------------------------+
The validation procedure is shown in Figure 2. Suppose the router has learned the SAV rules by SAV mechanisms and implemented them in the data plane. When a packet arrives at the router, the router will take the source address and the incoming interface of the packet as the input and look up the coresponding SAV rules. The final validity state that is either "valid" or "invalid" will be returned after the procedure.¶
Firstly, if the Mode 1 is enabled on the incoming interface, the packet will be only validated based on SAV rule 1 (interface-base prefix allowlist) -- according the principle that Mode 1 is the first priority and mutual exclusive with the other two modes.¶
And then, if the Mode 1 is not but Mode 2 is enabled on the incoming interface, the packet will be validated based on SAV rule 2 (interface-base prefix blocklist).¶
Mode 3 validation procedure will be gone through based on SAV rule 3 (prefix-based interface allowlist) only if: 1) the Mode 1 is not enabled on the incoming interface; 2) Mode 2 is not enabled on the incoming interface, or Mode 2 is enabled but the validaty state is valid; 3) Router-level Mode 3 is enabled.¶
SAV-enabled router +----------------------------------------------------------------------------+ | | packet-->|-->Mode 1?--Yes--> validating --> Validity State | | | on SAV rule 1 | | | | | +--No--> Mode 2?--Yes--> validating --> Validity State | | | on SAV rule 2 | | | | if Valid | | | | | | +--No-->Mode 3?<------------------------ + | | | | | +--Yes--> validating --> Validity State | | on SAV rule 3 | | | +----------------------------------------------------------------------------+
After doing validation, the router gets the validity state of the incoming packet. For the packet with invalid state, traffic handling policies should be taken on the packet. Simply forwarding or silently dropping may not well satisfy the requirements of operators in different scenarios. This section suggests to provide flexible traffic handling policies to validated packets just like FlowSpec ([RFC8955], [RFC8956]).¶
The followings are the traffic control policies that can be taken. One and only one of the policies will be chosen for an "invalid" validation result.¶
"Permit": Forward packets normally though the packets are considered invalid. This policy is useful when operators only want to monitor the status of source address spoofing in the network. The "Permit" policy can be taken together with the "Sample" policy.¶
"Discard": Drop packets directly, which is the common choose of existing SAV mechanisms.¶
"Rate limit": Enforce an upper bound of traffic rate (e.g., bps or pps) for mitigation of source address spoofing attacks. This policy is helpful while operators want do tentative filtering.¶
"Traffic redirect": Redirect the packets to the specified points (e.g., scrubbing centers) in the network for attack elimination.¶
There are also traffic monitor policies that are optional. One of the useful traffic monitor policies is:¶
"Sample": Capture the packets with a configurable sampling rate and reports them to remote servers (e.g., security analysis center). The sampled packets can be used for threat awareness and further analysis [I-D.cheng-savnet-proactive-defense-network]. "Sample" can be taken together with any one of the above policies. Note that, existing techniques like NetStream or NetFlow can be used for "Sample".¶
This document focuses on the general SAV capabilities. The generation of SAV rules is not discussed. There may be some security considerations for SAV generation, but it is not in the scope of this document.¶
The "Sample" policy pushes data to remote servers. This function can be achieved by existing techniques like NetStream or NetFlow. The "Sample" policy may induce same security considerations as these techniques, and the corresponding documents have discussed them.¶
This document includes no request to IANA.¶