]> Tsinghua University

Beijing China tolidan@tsinghua.edu.cn

Tsinghua University

Beijing China jianping@cernet.edu.cn

Huawei

Beijing China huangmingqing@huawei.com

Zhongguancun Laboratory

Beijing China lichen@zgclab.edu.cn

Huawei

Beijing China gengnan@huawei.com

Zhongguancun Laboratory

Beijing China liulb@zgclab.edu.cn

Tsinghua University

Beijing China qlc19@mails.tsinghua.edu.cn

General [REPLACE] Internet Engineering Task Force This document introduces an inter-domain SAVNET architecture for performing AS-level SAV and provides a comprehensive framework for guiding the design of inter-domain SAV mechanisms. The proposed architecture empowers ASes to generate SAV rules by sharing SAV-specific information between themselves, which can be used to generate more accurate and trustworthy SAV rules in a timely manner compared to the general information. During the incremental or partial deployment of SAV-specific information, it can utilize general information to generate SAV rules, if an AS's SAV-specific information is unavailable. Rather than delving into protocol extensions or implementations, this document primarily concentrates on proposing SAV-specific and general information and guiding how to utilize them to generate SAV rules. To this end, it also defines some architectural components and their relations.

Introduction Attacks based on source IP address spoofing, such as reflective DDoS and flooding attacks, continue to present significant challenges to Internet security. Mitigating these attacks in inter-domain networks requires effective source address validation (SAV). While BCP84 offers some SAV solutions, such as ACL-based ingress filtering and uRPF-based mechanisms, existing inter-domain SAV mechanisms have limitations in terms of validation accuracy and operational overhead in different scenarios . There are various existing general information from different sources including RPKI ROA objects and ASPA objects, RIB, FIB, and Internet Routing Registry (IRR) data, which can be used for inter-domain SAV. Generating SAV rules based on general information, however, cannot well satisfy the requirements for new inter-domain SAV mechanisms proposed in . As analyzed in , general information from RPKI ROA objects and ASPA objects can be used to infer the prefixes and their permissible incoming directions yet cannot be updated in a timely manner to adapt to the prefix or route changes, and the local routing information, which represents the general information from RIB or FIB, cannot deal with the asymmetric routing scenarios and may lead to improper blocks or improper permits, while IRR data do not update in a timely manner either and are not always accurate. Consequently, to address these issues, the inter-domain SAVNET architecture focuses on providing a comprehensive framework and guidelines for the design and implementation of new inter-domain SAV mechanisms. Inter-domain SAVNET architecture proposes SAV-specific information and uses it to generate SAV rules. SAV-specific information consists of prefixes and their corresponding legitimate incoming direction to enter an AS. Inter-domain SAVNET architecture can use it to generate more accurate SAV rules. In order to gather the SAV-specific information, a SAV-specific information communication mechanism would be developed for origination, processing, propagation, and termination of the messages which carry the SAV-specific information, and it can be implemented by a new protocol or extending an existing protocol. When the prefixes or routes change, it can update the SAV-specific information automatically in a timely manner. Also, the inter-domain SAVNET architecture will communicate the SAV-specific information over a secure connection between authenticated ASes. Moreover, during the incremental/partial deployment period of the SAV-specific information, the inter-domain SAVNET architecture can leverage the general information to generate SAV rules, if the SAV-specific information of an AS is unavailable. Multiple information sources may exist concurrently, to determine the one used for generating SAV rules, the inter-domain SAVNET architecture assigns priorities to the SAV-specific information and different general information and generates SAV rules using the SAV-related information with the highest-priority. SAV-specific information has the highest priority and the priorities of RPKI ROA objects and ASPA objects, RIB, FIB, and IRR data decrease in turn.

An example for illustrating the incentive of deploying inter-domain SAVNET architecture.

The inter-domain SAVNET architecture provides the incentive to deploy inter-domain SAV for operators. illustrates this using an example. P1 is the source prefix of AS 1, and AS 4 sends spoofing packets with P1 as source addresses to AS 3 through AS 2. Assume AS 4 does not deploy intra-domain SAV, these spoofing packets cannot be blocked by AS 4. Although AS 1 can deploy intra-domain SAV to block incoming packets which spoof the addresses of AS 1, these spoofing traffic from AS 4 to AS 3 do not go through AS 1, so they cannot be blocked by AS 1. Inter-domain SAVNET architecture can help in this scenario. If AS 1 and AS 2 deploy inter-domain SAVNET architecture, AS 2 knows that the packets with P1 as source addresses should come from AS 1, and the spoofing packets can thus be blocked by AS 2 since they come from the incorrect direction. Specifically, by proposing SAV-specific information and using it to generate SAV rules, the inter-domain SAVNET architecture gives more deployment incentive compared to existing inter-domain SAV mechanisms, which will be analyzed in . In addition, this document primarily proposes a high-level architecture for describing the communication flow of SAV-specific information and general information, guiding how to utilize the SAV-specific information and general information for generating SAV rules and deploy an inter-domain SAV mechanism between ASes. This document does not specify protocol extensions or implementations. Its purpose is to provide a conceptual framework and guidance for the design and development of inter-domain SAV mechanisms, allowing implementers to adapt and implement the architecture based on their specific requirements and network environments.

Requirements Language 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 when, and only when, they appear in all capitals, as shown here.

Terminology

SAV Rule:

The rule that indicates the validity of a specific source IP address or source IP prefix.

SAV Table:

The table or data structure that implements the SAV rules and is used for performing source address validation on the data plane.

SAV-specific Information:

The information that is specialized for SAV rule generation, includes the source prefixes and their legitimate incoming directions to enter an AS, and is gathered by the communication between ASes with the SAV-specific information communication mechanism.

SAV-specific Information Communication Mechanism:

The mechanism that is used to communicate SAV-specific information between ASes and can be implemented by a new protocol or an extension to an existing protocol.

Local Routing Information:

The information that is stored in ASBR's local RIB or FIB and can be used to generate SAV rules in addition to the routing purpose.

General Information:

The information that is not specialized for SAV but can be utilized to generate SAV rules, and is initially utilized for other purposes. Currently, the general information consists of the information from RPKI ROA objects and ASPA objects, local routing information, and the one from IRR data.

SAV-related Information:

The information that can be used to generate SAV rules and includes SAV-specific information and general information.

SAVNET Agent:

The agent within a SAVNET-adopting AS that is responsible for gathering SAV-related information and utilizing it to generate SAV rules.

SAV Information Base:

SAV information base is a table or data structure for storing SAV-related information collected from different SAV information sources and is a component within SAVNET agent.

SAV Information Base Manager:

SAV information base manager maitains the SAV-related information in the SAV information base and uses it to generate SAV rule accordingly, and is a component within SAVNET agent.

Improper Block:

The validation results that the packets with legitimate source addresses are blocked improperly due to inaccurate SAV rules.

Improper Permit:

The validation results that the packets with spoofed source addresses are permitted improperly due to inaccurate SAV rules.

Design Goals The inter-domain SAVNET architecture aims to improve SAV accuracy and facilitate partial deployment with low operational overhead, while guaranteeing convergence and providing security guarantees to the communicated information, which corresponds to the requirements for new inter-domain SAV mechanisms proposed in the inter-domain SAVNET architecture draft . The overall goal can be broken down into the following aspects:

• G1: The inter-domain SAVNET architecture should learn the real paths of source prefixes to any destination prefixes or permissible paths that can cover their real paths, and generate accurate SAV rules automatically based on the learned information to avoid improper blocks and reduce improper permits as much as possible.
• G2: The inter-domain SAVNET architecture should provide sufficient protection for the source prefixes of ASes that deploy it, even if only a portion of the Internet does the deployment.
• G3: The inter-domain SAVNET architecture should adapt to dynamic networks and asymmetric routing scenarios automatically.
• G4: The inter-domain SAVNET architecture should promptly detect the network changes and launch the convergence process in a timely manner, while reducing improper blocks and improper permits during the convergence process.
• G5: The inter-domain SAVNET architecture should provide security guarantees for the communicated SAV-specific information.

Other design goals, such as low operational overhead and easy implementation, are also very important and should be considered in specific protocols or protocol extensions.

Inter-domain SAVNET Architecture Overview

Inter-domain SAVNET architecture.

provides an overview of the inter-domain SAVNET architecture, showcasing an AS topology and the flow of SAV-related information among ASes. The topology captures the full spectrum of AS relationships in the Internet, displaying all peer ASes of AS X including customers, lateral peers, and providers and the existence of multiple physical links between ASes. Arrows in the figure indicate the direction of the corresponding SAV-related information from its source to AS X, such as gathering RPKI ROA objects and ASPA objects from RPKI cache server. The inter-domain SAVNET architecture conveys the SAV-related information through various mediums such as SAV-specific messages, BGP messages, RTR messages, and FTP messages. Based on the SAV-related information, AS X generates SAV rules. It is also worth noting that the inter-domain SAVNET architecture discusses AS-level inter-domain SAV. uses AS X as the representative to illustrate that what SAV-related information the SAVNET agent within AS X will collect and where the information is from. AS X has deployed SAVNET agent and can generate SAV rules to perform inter-domain SAV by consolidating the SAV-related information. It can obtain SAV-specific information from its customer AS which deploys SAVNET agent and local routing information originating from the BGP update messages of its neighbor ASes. Also, AS X can obtain RPKI ROA objects and ASPA objects from RPKI cache server and IRR data from IRR database. The inter-domain SAVNET architecture proposes SAV-specific information, which is more accurate and trustworthy than existing general information, and can update in a timely manner. SAV-specific information consists of prefixes and their legitimate incoming directions. The SAVNET agent communicates SAV-specific information between ASes via SAV-specific messages, when prefixes or routes change, it can launch SAV-specific messages timely to update SAV-specific information. Additionally, when SAVNET agent receives SAV-specific messages, it will validate whether the SAV-specific connections for communicating SAV-specific messages are authentic connections from authenticated ASes. Therefore, when SAV-specific information of an AS is available, SAVNET agent will use it to generate SAV rules. Furthermore, if the SAV-specific information is needed to communicate between ASes, a new SAV-specific information communication mechanism would be developed to exchange the SAV-specific messages between ASes which carry the SAV-specific information. It should define the data structure or format for communicating the SAV-specific information and the operations and timing for originating, processing, propagating, and terminating the SAV-specific messages. Also, it can be implemented by a new protocol or extending an existing protocol. The SAVNET agent should launch SAV-specific messages to adapt to the route changes in a timely manner. The SAV-specific information communication mechanism should handle route changes carefully to avoid improper blocks. The reasons for leading to improper blocks may include late detection of route changes, delayed message transmission, or packet losses. During the convergence process of the SAV-specific information communication mechanism, the inter-domain SAVNET architecture can use the information from RPKI ROA objects and ASPA objects to generate SAV rules until the convergence process is finished, since these information includes topological information and is more stable, and can thus avoid improper blocks. However, the detailed design of the SAV-specific information communication mechanism for dealing with route changes is outside the scope of this document. In the incremental/partial deployment stage of the inter-domain SAVNET architecture, when the SAV-specific information of some ASes is unavailable, SAVNET agent can leverage general information to generate SAV rules. If all these general information is available, it is recommended to use RPKI ROA objects and ASPA objects to generate SAV rules. Since compared to the local routing information and IRR data, they can provide authoritative prefixes and topological information and have less improper blocks. The systematic recommendations for the utilizations of SAV-related information and the corresponding rationale will be illustrated in . Regarding the security concerns, the inter-domain SAVNET architecture shares the similar security threats with BGP and can leverage existing BGP security mechanisms to enhance both session and content security.

SAVNET agent and SAV table within AS X in Figure 2.

displays the SAVNET agent and SAV table within AS X. The SAVNET agent can obtain the SAV-specific information and general information from various SAV information sources including SAV-specific messages from other ASes, RPKI cache server, and RIB or FIB as long as they are available. The SAV information base (SIB) within the SAVNET agent can store the SAV-specific information and general information and is maintained by the SIB manager. And SIB manager generates SAV rules based on the SIB and fills out the SAV table on the data plane. Moreover, the SIB can be managed by network operators using various methods such as YANG , Command-Line Interface (CLI), remote triggered black hole (RTBH) , and Flowspec . The detailed collection methods of the SAV-related information depend on the deployment and implementation of the inter-domain SAV mechanisms and are out of scope for this document. In the data plane, the packets coming from other ASes will be validated by the SAV table and only the packets which are permitted by the SAV table will be forwarded to the next hop. To achieve this, the router looks up each packet's source address in its local SAV table and gets one of three validity states: "Valid", "Invalid" or "Unknown". "Valid" means that there is a source prefix in SAV table covering the source address of the packet and the valid incoming interfaces covering the actual incoming interface of the packet. According to the SAV principle, "Valid" packets will be forwarded. "Invalid" means there is a source prefix in SAV table covering the source address, but the incoming interface of the packet does not match any valid incoming interface so that such packets will be dropped. "Unknown" means there is no source prefix in SAV table covering the source address. The packet with "unknown" addresses can be dropped or permitted, which depends on the choice of operators. The structure and detailed usage of SAV table can refer to .

SAV-related Information SAV-related information represents the information that can be used for SAV and consists of RPKI ROA objects and ASPA objects, local routing information, IRR data, and SAV-specific information. In the inter-domain SAVNET architecture, RPKI ROA objects and ASPA objects, local routing information, and IRR data are categorized into general information. In the future, if a new information source is created and can be used for SAV, but is not originally and specially used for SAV, its information can be categorized into general information. In other words, general information can also be considered as dual-use information.

General Information General information refers to the information that is not directly designed for SAV but can be utilized to generate SAV rules, and includes RPKI ROA objects and ASPA objects, local routing information, and IRR data.

RPKI ROA objects and ASPA Objects The RPKI ROA objects and ASPA objects are originally designed for the routing security purpose. RPKI ROA objects consists of {prefix, maximum length, origin AS} information and are originally used to mitigate the route origin hijacking, while RPKI ASPA objects consists of {ASN, Provider AS Set} information and are originally used to mitigate the route leaks. Both the objects are verified and authoritative. They are also stable and will not be updated frequently. Based on ASPA objects, the AS-level network topology can be constructed. And according to the ROA objects and the constructed AS-level topology information, an AS can learn all the permissible paths of the prefixes from its customer cone. Therefore, the prefixes and all its permissible incoming directions can be obtained. All the permissible incoming directions, however, do not only consist of the real incoming directions of the prefixes, but also the extra non-used incoming directions by the legitimate traffic, which would lead to improper permits. Additionally, according to a recent study , the process of updating RPKI information typically requires several minutes to an hour. This encompasses the addition or deletion of RPKI objects and the subsequent retrieval of updated information by ASes.

Local Routing Information The local routing information is originally used to guide the packet forwarding on each router and can be stored in the local RIB or FIB. It can be parsed from the BGP update messages communicated between ASes. Existing uRPF-based SAV mechanisms use the local routing information to generate SAV rules. As analyzed in , in the asymmetric routing scenarios, these mechanisms have accuracy problems and would lead to improper permits or improper blocks.

IRR Data The IRR data consist of ASes and their corresponding prefixes and can be used to augment the SAV table . However, only using IRR data for SAV would limit the functioning scope of SAV, in inter-domain networks, it may only be able to prevent spoofing by a stub AS. In addition, the IRR data are not always accurate.

SAV-specific Information SAV-specific information is the information that is specifically designed for SAV and consists of prefixes and their legitimate incoming directions to enter ASes. It can be contained in the SAV-specific messages which are communicated between ASes which deploy the inter-domain SAVNET architecture. When parsing the SAV-specific messages and obtaining the SAV-specific information, ASes can learn the prefixes and their legitimate incoming direction to enter themselves. Moreover, in the inter-domain SAVNET architecture, a SAV-specific information communication mechanism is used to communicate SAV-specific information between ASes and distribute the updated information to the relative ASes automatically in a timely manner once the prefixes or routes change.

Distinctions of Different SAV-related Information

The comprehensive comparasions between different SAV-related information.

shows the comprehensive comparasions between different SAV-related information when only using the corresponding information as the source to generate SAV rules and can help clarify their distinctions. Compared against general information, SAV-specific information is more accurate and trustworthy, while it can update the SAV rules in a timely manner to adapt to the prefix or route changes.

SAV Information Base

Priority ranking for the SAV information sources.

The SIB is managed by the SIB manager, which can consolidate SAV-related information from different sources. presents the priority ranking for the SAV-specific information and general information. Priority ranking from 1 to 5 represents high to low priority. Inter-domain SAVNET architecture uses SAV-related information from different sources based on their priorities. Once the SAV-specific information for a prefix is available within the SIB, the inter-domain SAVNET generates SAV rules based on SAV-specific information; otherwise, the inter-domain SAVNET generates SAV rules based on general information. The inter-domain SAVNET architecture assigns priorities to the information from different SAV information sources, and always generates the SAV rules using the information with the highest priority from all the available information. The priority ranking recommendation for different SAV information sources in is based on the accuracy, timeliness, trustness of the information from these sources. These properties determine that whether the requirements for new inter-domain SAV mechanisms proposed in can be well satisfied. SAV-specific information has higher priority than the general information, since it is specifically designed to carry more accurate SAV information, and can be updated in a timely manner to adapt to the prefix or route changes. The general information from RPKI ROA objects and ASPA objects, RIB, FIB, IRR data has different priorities, ranking 2, 3, 4, and 5, respectively. RPKI ROA objects and ASPA objects have higher priority than the one from RIB, FIB, and IRR data, this is because they can provide authoritative prefixes and topology information, which can be used to generate more accurate SAV rules. Also, they are more stable and can be used to reduce the risk of improper blocks during the convergence process of the network. Although the information source for RIB and FIB is the same, the RIB consists of more backup path information than the FIB, which can reduce improper blocks. IRR data have the lowest priority compared to others, since they are usually updated in a slower manner than the real network changes and not always correct.

An example of AS topology.

An example for the SAV information base of AS 4 in Figure 6.

We use the examples shown in and to introduce SIB and illustrate how to generate SAV rules based on the SIB. depicts an example of the SIB established in AS 4 displayed in . Each row of the SIB contains an index, prefix, incoming direction of the prefix, reltation between ASes, and the corresponding sources of the information. The incoming direction consists of customer, provider, and peer. For example, in , the row with index 0 indicates the incoming direction of P1 is AS 2 and the information source is SAV-specific information. Note that the same SAV-related information may have multiple sources and the SIB records them all, such as the row indexed 6. Moreover, SIB should be carefully implemented in the specific protocol or protocol extensions to avoid becoming a heavy burden of the router, and the similar optimization approaches used for the RIB may be applied. Recall that inter-domain SAVNET architecture generates SAV rules based on the SAV-related information in the SIB and their priorities. In addition, in the case of an AS's interfaces facing provider or lateral peer ASes where loose SAV rules are applicable, the inter-domain SAVNET architecture recommends to use blocklist at such directions to only block the prefixes that are sure not to come at these directions, while in the case of an AS's interfaces facing customer ASes that necessitate stricter SAV rules, the inter-domain SAVNET architecture recommends to use allowlist to only permit the prefixes that are allowed to come at these directions. Based on the above rules, taking the SIB in as an example to illustrate how the inter-domain SAVNET generates rules, AS 4 can conduct SAV as follows: SAV at the interfaces facing AS 3 blocks P1, P2, and P6 according to the rows indexed 0, 2, and 6 in the SIB, SAV at the interfaces facing AS 2 permits P1, P2, and P6 according to the rows indexed 0, 2, and 6 in the SIB, SAV at the interfaces facing AS 1 does not permit any prefixes according to the row indexed 0, 1, 6, and 7 in the SIB, and SAV at the interfaces facing AS 5 permits P5 according to the row indexed 5 in the SIB.

SAVNET Communication Mechanism

SAVNET communication mechanism for gathering SAV-related information from different SAV information sources.

SAV-specific information relies on the communication between SAVNET agents and general information can be from RPKI ROA objects and ASPA objects, RIB, FIB, and IRR data. Therefore, as illustrated in , the SAVNET agent needs to receive the SAV-related information from these SAV information sources. SAVNET agent also needs to accept the configurations from network operators for the management operations. Gathering these types of information relies on the SAVNET communication mechanism, which includes SAV-specific information communication mechanism, general information communication mechanism, and management mechanism.

SAV-specific Information Communication Mechanism

An example for exchanging SAV-specific information with SAV-specific information communication mechanism between AS 1 and AS X. | SAVNET | | | | Agent |<-|------------------------|--| Agent | | | +-------------+ | (P4, AS X) | +-------------+ | +------------------+ SAV-specific Messages +------------------+ ]]>

uses an example for exchanging SAV-specific information with SAV-specific messages between AS 1 and AS X. The SAV-specific information can be expressed as <Prefix, Incoming Direction> pairs, e.g., (P1, AS 2), (P6, AS 2), and (P4, AS X) in . The SAV-specific information can be exchanged between ASes via SAV-specific messages. SAV-specific messages are used to propagate or originate the SAV-specific information between ASes by the SAVNET agent. For an AS which initiates its own SAV-specific messages, the SAVNET agent within the AS can obtain incoming direction of its own prefixes to enter other ASes based on the local RIB and uses SAV-specific messages to carry the AS's prefixes to the corresponding ASes. When other ASes receive the SAV-specific messages, they parse the messages to obtain source prefixes and their corresponding incoming directions to enter these ASes. Additionally, if SAV-specific information is communicated between ASes, a new SAV-specific information communication mechanism would need to be developed to communicate it and can be implemented by a new protocol or extending an existing protocol. The SAV-specific information communication mechanism needs to define the data structure or format to communicate the SAV-specific messages and the operations and timing for originating, processing, propagating, and terminating the messages. If an extension to an existing protocol is used to exchange SAV-specific information, the corresponding existing protocol should not be affected. The SAVNET agent is the entity to support the SAV-specific communication mechanism. By parsing the SAV-specific messages, it obtains the prefixes and their incoming AS direction for maintaining the SIB. It is important to note that the SAVNET agent within an AS has the capability to establish connections with multiple SAVNET agents within different ASes, relying on either manual configurations by operators or an automatic mechanism. In addition, SAVNET agents should validate the authenticity of the connection for communicating the SAV-specific information to verify whether the SAV-specific information is provided over a secure connection with an authenticated AS. The need for a SAV-specific communication mechanism arises from the facts that the SAV-specific information needs to be obtained and communicated between ASes. Different from the general information such as routing information from the RIB, there are no existing mechanism which can support the perception and communication of SAV-specific information between ASes. Hence, a SAV-specific communication mechanism is needed to provide a medium and set of rules to establish communication between different ASes for the exchange of SAV-specific information. Furthermore, an AS needs to assemble its source prefixes into the SAV-specific messages. In order to obtain all the source prefixes of an AS, the inter-domain SAVNET architecture can communicate with the intra-domain SAVNET architecture to obtain all the prefixes belonging to an AS. The preferred AS paths of an AS may change over time due to route changes or network failures. The SAVNET agent should launch SAV-specific messages to adapt to the route changes in a timely manner. The SAV-specific information communication mechanism should handle route changes carefully to avoid improper blocks. The reasons for leading to improper blocks may include late detection of route changes, delayed message transmission, or packet losses. However, the detailed design of SAV-specific information communication mechanism for dealing with route changes is outside the scope of this document.

General Information Communication Mechanism The general information communication mechanism is used for communicating routing information between ASes, obtaining RPKI ROA objects and ASPA objects from RPKI cache servers, and obtaining the information about ASes and their prefixes from IRR databases. The general communication mechanism can be implemented by using existing protocols for collecting the relative information, such as BGP, RTR , and FTP .

Management Mechanism The primary purpose of the management mechanism is to deliver manual configurations of network operators. Examples of the management configurations include, but are not limited to:

• SAVNET configurations using YANG, CLI, RTBH, or Flowspec.
• SAVNET operation.
• Inter-domain SAVNET provisioning.

Note that the configuration information can be delivered at any time and requires reliable delivery for the management mechanism implementation. Additionally, the management mechanism can carry telemetry information, such as metrics pertaining to forwarding performance, the count of spoofing packets and discarded packets, provided that the inter-domain SAVNET has access to such data. It can include information regarding the prefixes associated with the spoofing traffic, as observed until the most recent time.

Use Cases This section utilizes the sample use cases to showcase that the inter-domain SAVNET architecture can improve the validation accuracy in the scenarios of limited propagation of prefixes, hidden prefixes, reflection attacks, and direct attacks, compared to existing SAV mechanisms, which are also utilized for the gap analysis of existing inter-domain SAV mechanisms in . In the following, these use cases are discussed for SAV at customer interfaces and SAV at provider/peer interfaces, respectively.

SAV at Customer Interfaces In order to prevent the source address spoofing, operators can enable ACL-based ingress filtering, source-based RTBH filtering, and/or uRPF-based mechanisms at customer interfaces, namely Strict uRPF, FP-uRPF, VRF uRPF, or EFP-uRPF . However, as analyzed in , uRPF-based mechanisms may lead to false positives in two inter-domain scenarios: limited propagation of prefixes and hidden prefixes, or may lead to false negatives in the scenarios of source address spoofing attacks within a customer cone, while ACL-based ingress filtering and source-based RTBH filtering need to update SAV rules in a timely manner and lead to high operational overhead. The following showcases that the inter-domain SAVNET architecture can avoid false positives and false negatives in these scenarios.

Limited Propagation of Prefixes

Limited propagation of prefixes caused by NO_EXPORT.

presents a scenario where the limited propagation of prefixes occurs due to the NO_EXPORT community attribute. In this scenario, AS 1 is a customer of AS 2, AS 2 is a customer of AS 4, AS 4 is a customer of AS 3, and AS 5 is a customer of both AS 3 and AS 4. The relationship between AS 1 and AS 4 can be either customer-to-provider (C2P) or peer-to-peer (P2P). AS 1 advertises prefixes P1 to AS 2 and adds the NO_EXPORT community attribute to the BGP advertisement sent to AS 2, preventing AS 2 from further propagating the route for prefix P1 to AS 4. Similarly, AS 1 adds the NO_EXPORT community attribute to the BGP advertisement sent to AS 4, resulting in AS 4 not propagating the route for prefix P6 to AS 3. Consequently, AS 4 only learns the route for prefix P1 from AS 1 in this scenario. Suppose AS 1 and AS 4 have deployed inter-domain SAV while other ASes have not, and AS 4 has deployed EFP-uRPF at its customer interfaces. In this scenario, existing uRPF-based SAV mechanisms would block the traffic with P1 as source addresses improperly, and thus suffer from the problem of false positives . If the inter-domain SAVNET architecture is deployed, AS 1 can communicate the SAV-specific information to AS 4 and AS 4 will be aware that the traffic with P1 as source addresses can arrive at the interfaces facing AS 1 and AS 2. As a result, the false positive problem can be avoided.

Hidden Prefixes

A Direct Server Return (DSR) scenario.

illustrates a direct server return (DSR) scenario where the anycast IP prefix P3 is only advertised by AS 3 through BGP. In this example, AS 3 is the provider of AS 4 and AS 5, AS 4 is the provider of AS 1, AS 2, and AS 5, and AS 2 is the provider of AS 1. AS 1 and AS 4 have deployed inter-domain SAV, while other ASes have not. When users in AS 2 send requests to the anycast destination IP, the forwarding path is AS 2->AS 4->AS 3. The anycast servers in AS 3 receive the requests and tunnel them to the edge servers in AS 1. Finally, the edge servers send the content to the users with source addresses in prefix P3. The reverse forwarding path is AS 1->AS 4->AS 2. In this scenario, existing uRPF-based mechanisms will improperly block the legitimate response packets from AS 1 at the customer interface of AS 4 facing AS 1 . In contrast, if the inter-domain SAVNET architecture is deployed, AS 1 can communicate the SAV-specific information to AS 4 and AS 4 will be aware that the traffic with P3 as source addresses can arrive at the interfaces facing AS 1 and AS 3. As a result, the legitimate response packets with P3 as source addresses from AS 1 can be allowed and the false positive problem can be avoided.

Reflection Attacks

A scenario of reflection attacks by source address spoofing within a customer cone.

depicts the scenario of reflection attacks by source address spoofing within a customer cone. The reflection attack by source address spoofing takes place within AS 4's customer cone, where the attacker spoofs the victim's IP address (P1) and sends requests to servers' IP address (P5) that are designed to respond to such requests. As a result, the server sends overwhelming responses back to the victim, thereby exhausting its network resources. The arrows in illustrate the commercial relationships between ASes. AS 3 serves as the provider for AS 4 and AS 5, while AS 4 acts as the provider for AS 1, AS 2, and AS 5. Additionally, AS 2 is the provider for AS 1. Suppose AS 1 and AS 4 have deployed inter-domain SAV, while the other ASes have not. In this scenario, EFP-uRPF with algorithm A/B will improperly permit the spoofing attacks originating from AS 2 . If the inter-domain SAVNET architecture is deployed, AS 1 can communicate the SAV-specific information to AS 4 and AS 4 will be aware that the traffic with P1 as source addresses can only arrive at the interface facing AS 1. Therefore, at the interface of AS 4 facing AS 2, the spoofing traffic can be blocked.

Direct Attacks

A scenario of the direct attacks by source address spoofing within a customer cone.

portrays a scenario of direct attacks by source address spoofing within a customer cone and is used to analyze the gaps of uRPF-based mechanisms below. The direct attack by source address spoofing takes place within AS 4's customer cone, where the attacker spoofs a source address (P5) and directly targets the victim's IP address (P1), overwhelming its network resources. The arrows in illustrate the commercial relationships between ASes. AS 3 serves as the provider for AS 4 and AS 5, while AS 4 acts as the provider for AS 1, AS 2, and AS 5. Additionally, AS 2 is the provider for AS 1. Suppose AS 1 and AS 4 have deployed inter-domain SAV, while the other ASes have not. In this scenario, EFP-uRPF with algorithm A/B will improperly permit the spoofing attacks . If the inter-domain SAVNET architecture is deployed, AS 5 can communicate the SAV-specific information to AS 4 and AS 4 will be aware that the traffic with P5 as source addresses can arrive at the interface facing AS 3 and AS 5. Therefore, at the interface of AS 4 facing AS 2, the spoofing traffic can be blocked.

SAV at Provider/Peer Interfaces In order to prevent packets with spoofed source addresses from the provider/peer AS, ACL-based ingress filtering, Loose uRPF, and/or source-based RTBH filtering can be deployed . exposes the limitations of ACL-based ingress filtering, source-based RTBH filtering, and Loose uRPF for SAV at provider/peer interfaces in scenarios of source address spoofing attacks from provider/peer AS. The source address spoofing attacks from provider/peer AS include reflection attacks from provider/peer AS and direct attacks from provider/peer AS. The following showcases that the inter-domain SAVNET architecture can avoid false negatives in these scenarios.

Reflection Attacks

A scenario of reflection attacks by source address spoofing from provider/peer AS.

depicts the scenario of reflection attacks by source address spoofing from provider/peer AS. In this case, the attacker spoofs the victim's IP address (P1) and sends requests to servers' IP address (P2) that respond to such requests. The servers then send overwhelming responses back to the victim, exhausting its network resources. The arrows in represent the commercial relationships between ASes. AS 3 acts as the provider or lateral peer of AS 4 and the provider for AS 5, while AS 4 serves as the provider for AS 1, AS 2, and AS 5. Additionally, AS 2 is the provider for AS 1. Suppose AS 1 and AS 4 have deployed inter-domain SAV, while the other ASes have not. Both ACL-based ingress filtering and source-based RTBH filtering will induce additional operational overhead, and Loose uRPF may improperly permit spoofed packets . If the inter-domain SAVNET architecture is deployed, AS 1 can communicate the SAV-specific information to AS 4 and AS 4 will be aware that the traffic with P1 as source addresses can arrive at the interface facing AS 1 and AS 2. Therefore, at the interface of AS 4 facing AS 3, the spoofing traffic can be blocked.

Direct Attacks

A scenario of direct attacks by source address spoofing from provider/peer AS.

showcases a scenario of direct attack by source address spoofing from provider/peer AS. In this case, the attacker spoofs another source address (P2) and directly targets the victim's IP address (P1), overwhelming its network resources. The arrows in represent the commercial relationships between ASes. AS 3 acts as the provider or lateral peer of AS 4 and the provider for AS 5, while AS 4 serves as the provider for AS 1, AS 2, and AS 5. Additionally, AS 2 is the provider for AS 1. Suppose AS 1 and AS 4 have deployed inter-domain SAV, while the other ASes have not. Also, in this scenario, both ACL-based ingress filtering and source-based RTBH filtering will induce additional operational overhead, and Loose uRPF may improperly permit spoofed packets . If the inter-domain SAVNET architecture is deployed, AS 2 can communicate the SAV-specific information to AS 4 and AS 4 will be aware that the traffic with P2 as source addresses can only arrive at the interface facing AS 2. Therefore, at the interface of AS 4 facing AS 3, the spoofing traffic can be blocked.

Partial/Incremental Deployment Considerations The inter-domain SAVNET architecture MUST ensure support for partial/incremental deployment as it is not feasible to deploy it simultaneously in all ASes. The partial/incremental deployment of the inter-domain SAVNET architecture consists of different aspects, which include the partial/incremental deployment of the architecture and the partial/incremental deployment of the information sources. Within the architecture, the general information like the prefixes and topological information from RPKI ROA Objects and ASPA Objects and the routing information from the RIB can be obtained locally when the corresponding sources are available. Even when both SAV-specific Information and the information from RPKI ROA Objects and ASPA Objects are not available, the routing information from the RIB can be used to generate SAV rules. Furthermore, it is not mandatory for all ASes to deploy SAVNET agents for SAV-specific Information. Instead, a SAVNET agent should be able to effortlessly establish a logical neighboring relationship with another AS that has deployed a SAVNET agent. The connections for communicating SAV-specific Information can be achieved by manual configurations set by operators or an automatic neighbor discovery mechanism. This flexibility enables the architecture to accommodate varying degrees of deployment, promoting interoperability and collaboration among participating ASes. During the partial/incremental deployment of SAVNET agent, the SAV-specific Information for the ASes which do not deploy SAVNET agent can not be obtained. To protect the prefixes of these ASes, inter-domain SAVNET architecture can use the SAV-related information from the general information in the SIB to generate SAV rules. At least, the routing information from the RIB can be always available in the SIB. As more ASes adopt the inter-domain SAVNET architecture, the "deployed area" expands, thereby increasing the collective defense capability against source address spoofing. Furthermore, if multiple "deployed areas" can be logically interconnected across "non-deployed areas", these interconnected "deployed areas" can form a logical alliance, providing enhanced protection against address spoofing. Especially, along with more ASes deploy SAVNET agent and support the communication of SAV-specific information, the generated SAV rules of the inter-domain SAVNET architecture to protect these ASes will become more accurate, as well as enhancing the protection capability against source address spoofing for the inter-domain SAVNET architecture. In addition, releasing the SAV functions of the inter-domain SAVNET architecture incrementally is one potential way to reduce the deployment risks and can be considered in its deployment by network operators:

• First, the inter-domain SAVNET can only do the measurement in the data plane and do not take any other actions. Based on the measurement data, the operators can evaluate the effect of the inter-domain SAVNET on the legitimate traffic, including validation accuracy and forwarding performance, as well as the operational overhead.
• Second, the inter-domain SAVNET can open the function to limit the rate of the traffic that is justified as spoofing traffic. The operators can further evaluate the effect of the inter-domain SAVNET on the legitimate traffic and spoofing traffic, such as limiting the rate of all the spoofing traffic without hurting the legitimate traffic.
• Third, when the validation accuracy, forwarding performance, and operational overhead have been verified on a large scale by the live network, the inter-domain SAVNET can open the function to directly block the spoofing traffic that is justified by the SAV table in the data plane.

Convergence Considerations Convergence issues SHOULD be carefully considered in inter-domain SAV mechanisms due to the dynamic nature of the Internet. Internet routes undergo continuous changes, and SAV rules MUST proactively adapt to these changes, such as prefix and topology changes, in order to prevent false positives and reduce false negatives. To effectively track these changes, the SIM should promptly collect SAV-related information from various SAV information sources and consolidate them in a timely manner. In particular, it is essential for the SAVNET agents to proactively communicate the changes of the SAV-specific Information between ASes and adapt to route changes promptly. However, during the routing convergence process, the traffic paths of the source prefixes can undergo rapid changes within a short period. The changes of the SAV-specific Information may not be communicated in time between ASes to update SAV rules, false positives or false negatives may happen. Such inaccurate validation is caused by the delays in communicating SAV-specific Information between ASes, which occur due to the factors like packet losses, unpredictable network latencies, or message processing latencies. The design of the SAV-specific communication mechanism should consider these issues to reduce the inaccurate validation. Besides, for the inter-domain SAVNET architecture, the potential ways to deal with the inaccurate validation issues during the convergence of the SAV-specific communication mechanism is to consider using the information from RPKI ROA objects and ASPA objects to generate SAV rules until the convergence process of the SAV-specific communication mechanism is finished, since these information is more stable and can help avoid false positives, and thus avoiding the impact to the legitimate traffic.

Manageability Considerations It is crucial to consider the operations and management aspects of SAV information sources, the SAV-specific communication mechanism, SIB, SIM, and SAV table in the inter-domain SAVNET architecture. The following guidelines should be followed for their effective management: First, management interoperability should be supported across devices from different vendors or different releases of the same product, based on a unified data model such as YANG . This is essential because the Internet comprises devices from various vendors and different product releases that coexist simultaneously. Second, scalable operation and management methods such as NETCONF and syslog protocol should be supported. This is important as an AS may have hundreds or thousands of border routers that require efficient operation and management. Third, management operations, including default initial configuration, alarm and exception reporting, logging, performance monitoring and reporting for the control plane and data plane, as well as debugging, should be designed and implemented in the protocols or protocol extensions. These operations can be performed either locally or remotely, based on the operational requirements. By adhering to these rules, the management of SAV information sources and related components can be effectively carried out, ensuring interoperability, scalability, and efficient operations and management of the inter-domain SAVNET architecture.

Security Considerations In the inter-domain SAVNET architecture, the SAVNET agent plays a crucial role in generating and disseminating SAV-specific messages across different ASes. To safeguard against the potential risks posed by a malicious AS generating incorrect or forged SAV-specific messages, it is important for the SAVNET agents to employ security authentication measures for each received SAV-specific Message. The majour security threats faced by inter-domain SAVNET can be categorized into two aspects: session security and content security. Session security pertains to verifying the identities of both parties involved in a session and ensuring the integrity of the session content. Content security, on the other hand, focuses on verifying the authenticity and reliability of the session content, thereby enabling the identification of forged SAV-specific Messages. The threats to session security include:

• Session identity impersonation: This occurs when a malicious router deceitfully poses as a legitimate peer router to establish a session with the targeted router. By impersonating another router, the malicious entity can gain unauthorized access and potentially manipulate or disrupt the communication between the legitimate routers.
• Session integrity destruction: In this scenario, a malicious intermediate router situated between two peering routers intentionally tampers with or destroys the content of the relayed SAV-specific Message. By interfering with the integrity of the session content, the attacker can disrupt the reliable transmission of information, potentially leading to miscommunication or inaccurate SAV-related data being propagated.

The threats to content security include:

• Message alteration: A malicious router has the ability to manipulate or forge any portion of a SAV-specific message. For example, the attacker may employ techniques such as using a spoofed Autonomous System Number (ASN) or modifying the AS Path information within the message. By tampering with the content, the attacker can potentially introduce inaccuracies or deceive the receiving ASes, compromising the integrity and reliability of the SAV-related information.
• Message injection: A malicious router injects a seemingly "legitimate" SAV-specific message into the communication stream and directs it to the corresponding next-hop AS. This type of attack can be likened to a replay attack, where the attacker attempts to retransmit previously captured or fabricated messages to manipulate the behavior or decisions of the receiving ASes. The injected message may contain malicious instructions or false information, leading to incorrect SAV rule generation or improper validation.
• Path deviation: A malicious router intentionally diverts a SAV-specific Message to an incorrect next-hop AS, contrary to the expected path defined by the AS Path. By deviating from the intended routing path, the attacker can disrupt the proper dissemination of SAV-related information and introduce inconsistencies or conflicts in the validation process. This can undermine the effectiveness and accuracy of source address validation within the inter-domain SAVNET architecture.

Overall, inter-domain SAVNET shares similar security threats with BGP and can leverage existing BGP security mechanisms to enhance both session and content security. Session security can be enhanced by employing session authentication mechanisms used in BGP. Similarly, content security can benefit from the deployment of existing BGP security mechanisms like RPKI, BGPsec, and ASPA. While these mechanisms can address content security threats, their widespread deployment is crucial. Until then, it is necessary to develop an independent security mechanism specifically designed for inter-domain SAVNET. One potential approach is for each origin AS to calculate a digital signature for each AS path and include these digital signatures within the SAV-specific messages. Upon receiving a SAV-specific Message, the SAVNET agent can verify the digital signature to ascertain the message's authenticity. Furthermore, it is worth noting that the information channel of the inter-domain SAVNET architecture may need to operate over a network link that is currently under a source address spoofing attack. As a result, it may experience severe packet loss and high latency due to the ongoing attack, and the implementation of the information channel should ensure uninterrupted communication. Detailed security designs and considerations will be addressed in a separate draft, ensuring the robust security of inter-domain SAVNET.

Privacy Considerations TBD

IANA Considerations This document has no IANA requirements.

Scope and Assumptions In this architecture, the choice of protocols used for communication between the SIM and different SAV information sources is not limited. The inter-domain SAVNET architecture presents considerations on how to consolidate SAV-related information from various sources to generate SAV rules and perform SAV using the SAV table in the dataplane. The detailed design and implementation for SAV rule generation and SAV execution depend on the specific inter-domain SAV mechanisms employed. This document does not cover administrative or business agreements that may be established between the involved inter-domain SAVNET parties. These considerations are beyond the scope of this document. However, it is assumed that authentication and authorization mechanisms can be implemented to ensure that only authorized ASes can communicate SAV-related information. This document makes the following assumptions:

• All ASes where the inter-domain SAVNET is deployed are assumed to provide the necessary connectivity between SAVNET agent and any intermediate network elements. However, the architecture does not impose any specific limitations on the form or nature of this connectivity.
• Congestion and resource exhaustion can occur at various points in the inter-domain networks. Hence, in general, network conditions should be assumed to be hostile. The inter-domain SAVNET architecture must be capable of functioning reliably under all circumstances, including scenarios where the paths for delivering SAV-related information are severely impaired. It is crucial to design the inter-domain SAVNET system with a high level of resilience, particularly under extremely hostile network conditions. The architecture should ensure uninterrupted communication between inter-domain SAVNET agents, even when data-plane traffic saturates the link.
• The inter-domain SAVNET architecture does not impose rigid requirements for the SAV information sources that can be used to generate SAV rules. Similarly, it does not dictate strict rules on how to utilize the SAV-related information from diverse sources or perform SAV in the dataplane. Network operators have the flexibility to choose their approaches to generate SAV rules and perform SAV based on their specific requirements and preferences. Operators can either follow the recommendations outlined in the inter-domain SAVNET architecture or manually specify the rules for governing the use of SAV-related information, the generation of SAV rules, and the execution of SAV in the dataplane.
• The inter-domain SAVNET architecture does not impose restrictions on the selection of the local AS with which AS to communicate SAV-specific Information. The ASes have the flexibility to establish connections for SAV-specific communication based on the manual configurations set by operators or other automatic mechanisms.
• The inter-domain SAVNET architecture provides the flexibility to accommodate Quality-of-Service (QoS) policy agreements between SAVNET-enabled ASes or local QoS prioritization measures, but it does not make assumptions about their presence. These agreements or prioritization efforts are aimed at ensuring the reliable delivery of SAV-specific Information between SAVNET agents. It is important to note that QoS is considered as an operational consideration rather than a functional component of the inter-domain SAVNET architecture.
• The SAVNET communication mechanisms are loosely coupled and are used for communicating or gathering SAV-related information, and how the inter-domain SAVNET synchronizes the management and operation configurations is out of scope of this document.

Contributors Igor Lubashev
Akamai Technologies
145 Broadway
Cambridge, MA, 02142
United States of America
Email: ilubashe@akamai.com Many thanks to Igor Lubashev for the significantly helpful revision suggestions.

References Normative References RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. BCP 38, RFC 2827, is designed to limit the impact of distributed denial of service attacks, by denying traffic with spoofed addresses access to the network, and to help ensure that traffic is traceable to its correct source network. As a side effect of protecting the Internet against such attacks, the network implementing the solution also protects itself from this and other attacks, such as spoofed management access to networking equipment. There are cases when this may create problems, e.g., with multihoming. This document describes the current ingress filtering operational mechanisms, examines generic issues related to ingress filtering, and delves into the effects on multihoming in particular. This memo updates RFC 2827. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. This document identifies a need for and proposes improvement of the unicast Reverse Path Forwarding (uRPF) techniques (see RFC 3704) for detection and mitigation of source address spoofing (see BCP 38). Strict uRPF is inflexible about directionality, the loose uRPF is oblivious to directionality, and the current feasible-path uRPF attempts to strike a balance between the two (see RFC 3704). However, as shown in this document, the existing feasible-path uRPF still has shortcomings. This document describes enhanced feasible-path uRPF (EFP-uRPF) techniques that are more flexible (in a meaningful way) about directionality than the feasible-path uRPF (RFC 3704). The proposed EFP-uRPF methods aim to significantly reduce false positives regarding invalid detection in source address validation (SAV). Hence, they can potentially alleviate ISPs' concerns about the possibility of disrupting service for their customers and encourage greater deployment of uRPF techniques. This document updates RFC 3704. YANG is a data modeling language used to model configuration and state data manipulated by the Network Configuration Protocol (NETCONF), NETCONF remote procedure calls, and NETCONF notifications. [STANDARDS-TRACK] The Network Configuration Protocol (NETCONF) defined in this document provides mechanisms to install, manipulate, and delete the configuration of network devices. It uses an Extensible Markup Language (XML)-based data encoding for the configuration data as well as the protocol messages. The NETCONF protocol operations are realized as remote procedure calls (RPCs). This document obsoletes RFC 4741. [STANDARDS-TRACK] This document describes the syslog protocol, which is used to convey event notification messages. This protocol utilizes a layered architecture, which allows the use of any number of transport protocols for transmission of syslog messages. It also provides a message format that allows vendor-specific extensions to be provided in a structured way. This document has been written with the original design goals for traditional syslog in mind. The need for a new layered specification has arisen because standardization efforts for reliable and secure syslog extensions suffer from the lack of a Standards-Track and transport-independent RFC. Without this document, each other standard needs to define its own syslog packet format and transport mechanism, which over time will introduce subtle compatibility issues. This document tries to provide a foundation that syslog extensions can build on. This layered architecture approach also provides a solid basis that allows code to be written once for each syslog feature rather than once for each transport. [STANDARDS-TRACK] In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. Informative References Remote Triggered Black Hole (RTBH) filtering is a popular and effective technique for the mitigation of denial-of-service attacks. This document expands upon destination-based RTBH filtering by outlining a method to enable filtering by source address as well. This memo provides information for the Internet community. This document defines a Border Gateway Protocol Network Layer Reachability Information (BGP NLRI) encoding format that can be used to distribute (intra-domain and inter-domain) traffic Flow Specifications for IPv4 unicast and IPv4 BGP/MPLS VPN services. This allows the routing system to propagate information regarding more specific components of the traffic aggregate defined by an IP destination prefix. It also specifies BGP Extended Community encoding formats, which can be used to propagate Traffic Filtering Actions along with the Flow Specification NLRI. Those Traffic Filtering Actions encode actions a routing system can take if the packet matches the Flow Specification. This document obsoletes both RFC 5575 and RFC 7674. In order to verifiably validate the origin Autonomous Systems and Autonomous System Paths of BGP announcements, routers need a simple but reliable mechanism to receive Resource Public Key Infrastructure (RFC 6480) prefix origin data and router keys from a trusted cache. This document describes a protocol to deliver them. This document describes version 1 of the RPKI-Router protocol. RFC 6810 describes version 0. This document updates RFC 6810. This memo is the official specification of the File Transfer Protocol (FTP) for the DARPA Internet community. The primary intent is to clarify and correct the documentation of the FTP specification, not to change the protocol. The following new optional commands are included in this edition of the specification: Change to Parent Directory (CDUP), Structure Mount (SMNT), Store Unique (STOU), Remove Directory (RMD), Make Directory (MKD), Print Directory (PWD), and System (SYST). Note that this specification is compatible with the previous edition. MANRS NIST ISOC n.d.

Acknowledgements Many thanks to Alvaro Retana, Kotikalapudi Sriram, Rüdiger Volk, Xueyan Song, Ben Maddison, Jared Mauch, Joel Halpern, Aijun Wang, Jeffrey Haas, Xiangqing Chang, Changwang Lin, Mingxing Liu, Zhen Tan, Yuanyuan Zhang, Yangyang Wang, Antoin Verschuren etc. for their valuable comments on this document.