SAVI J. Bi Internet-Draft B. Liu Intended status: Informational Tsinghua Univ. Expires: November 10, 2017 May 9, 2017 Problem Statement of SAVI Beyond the First Hop draft-bi-savi-problem-13 Abstract IETF Source Address Validation Improvements (SAVI) working group is chartered for source address validation within the first hop from end hosts, i.e., preventing a node from spoofing the IP source address of another node in the same IP link. However, since SAVI requires the edge routers or switches to be upgraded, the deployment of SAVI will need a long time. During this transition period, some source address validation techniques beyond the first hop (SAVI-BF) may be needed to complement SAVI and protect the networks from spoofing based attacks. In this document, we first propose three desired features of the SAVI-BF techniques. Then we analyze the problems of the current SAVI-BF technique, ingress filtering. Finally, we discuss the directions that we can explore to improve SAVI-BF. Status of This Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." This Internet-Draft will expire on November 10, 2017. Copyright Notice Copyright (c) 2017 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 Bi & Liu Expires November 10, 2017 [Page 1] Internet-Draft SAVI Problem Beyond First Hop May 2017 (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. This document may contain material from IETF Documents or IETF Contributions published or made publicly available before November 10, 2008. The person(s) controlling the copyright in some of this material may not have granted the IETF Trust the right to allow modifications of such material outside the IETF Standards Process. Without obtaining an adequate license from the person(s) controlling the copyright in such materials, this document may not be modified outside the IETF Standards Process, and derivative works of it may not be created outside the IETF Standards Process, except to format it for publication as an RFC or to translate it into languages other than English. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 2. Desired Features of SAVI-BF Techniques . . . . . . . . . . . 3 2.1. High Deployment Incentives . . . . . . . . . . . . . . . 3 2.2. Low Operational Risks . . . . . . . . . . . . . . . . . . 3 2.3. Low Cost . . . . . . . . . . . . . . . . . . . . . . . . 4 3. Problems of Ingress Filtering . . . . . . . . . . . . . . . . 4 3.1. IAL . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 3.2. Strict/feasible RPF . . . . . . . . . . . . . . . . . . . 5 3.3. Loose RPF* . . . . . . . . . . . . . . . . . . . . . . . 6 3.4. Other Reasons . . . . . . . . . . . . . . . . . . . . . . 6 4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 7 4.1. Path based Techniques . . . . . . . . . . . . . . . . . . 7 4.2. End-to-end based Techniques . . . . . . . . . . . . . . . 7 4.3. Non-technical Proposals . . . . . . . . . . . . . . . . . 8 5. Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . 8 6. Informative References . . . . . . . . . . . . . . . . . . . 8 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 9 1. Introduction IETF Source Address Validation Improvements (SAVI) working group is chartered for source address validation within the first hop from the end hosts, so as to prevent a node from spoofing the IP source address of another node in the same IP link. However, since SAVI requires the edge routers or switches to be upgraded, the deployment of SAVI will need a long time. During this transition period, some Bi & Liu Expires November 10, 2017 [Page 2] Internet-Draft SAVI Problem Beyond First Hop May 2017 source address validation techniques beyond the first hop (SAVI-BF) may be needed to complement SAVI, so as to protect the networks from spoofing based attacks, which are prevalent DDoS attacks on the current Internet [Ground-Truth] [ARBOR-2010] [NANOG-Helpless] [DrDoS-300Gbps]. In this document, we first propose three desired features of SAVI-BF techniques. The first desired feature is high deployment incentives, i.e., by deploying a technique, an ISP should significantly increase its ability to protect its network from spoofing based attacks. The second one is low operational risks. If a technique may improperly drop legitimate packets (so called false positives), it introduces new risks to the network operation and management. It is desired that the false positives (FP) is as low as possible. The third one low cost. It is desired that the technique requires minimum deployment investment and operational cost. We evaluate ingress filtering [BCP38] [BCP84], the best current practice in for SAVI-BF, against these three features. Recent measurement shows that, the deployment of ingress filtering has not been improved over four years because the ISPs do not have incentives to deploy it [Efficacy], and sophisticated attackers exploit the spoofable networks to launch attacks. We discuss the reasons why ingress filtering is still insufficiently applied by the ISPs despite that it has long been available in modern routers. Finally, we discuss the directions that we can explore to improve SAVI-BF. We briefly survey two categories of SAVI-BF proposals, the path based techniques and the end-to-end based techniques. We discuss their requirements, advantages and disadvantages. 2. Desired Features of SAVI-BF Techniques 2.1. High Deployment Incentives To motivate an ISP to deploy a technique, it is desirable that the technique can generate additional benefit for the ISP. In terms of a SAVI-BF technique, it should provide the ISP with additional protection from spoofing based attacks. Specifically, it should significantly decrease the volume of the spoofing based attacks targeting the ISP, or the number of networks where these attacks can be successfully launched, or the number of source addresses or destination addresses that can be used in these attacks. 2.2. Low Operational Risks If a technique may improperly drop legitimate packets, so called false positives (FP), it introduces new operational risks. Bi & Liu Expires November 10, 2017 [Page 3] Internet-Draft SAVI Problem Beyond First Hop May 2017 Apparently, it is desired that the FP is as low as possible. The higher the FP, the more limited its application scope. For example, if the attack is not very severe, the network administrator would not apply a technique with high FP, since its operation risks may even surpass the loss caused by the attack. 2.3. Low Cost It is desired that the technique requires minimum investment on the device and system upgrading. It should also adapt to network dynamics rather than require manual intervention, which is costly, slow, and often the source of errors (misconfiguration). 3. Problems of Ingress Filtering Ingress filtering is the best current practice for SAVI-BF. Ingress filtering was proposed in 2000 [BCP38] , and updated in 2004 [BCP84]. Ingress access lists (IALs) and unicast reverse path forwarding (uRPF) are two general ways to implement ingress filtering. An IAL is a filter that checks the source address of every packet received on a network interface against a list of acceptable prefixes, dropping any packet that does not match the filter. IALs are typically maintained manually. Upon network dynamics (topology change or routing change), the IALs should be updated accordingly to avoid dropping legitimate packets. uRPF is an automated tool which can adapt with the network dynamics. On the receipt of a packet, uRPF checks its source address against the forwarding table or routing table for validation. uRPF has four variants, strict RPF, feasible RPF, loose RPF and loose RPF ignoring default routes. Readers are referred to [BCP84] for the details of these variants. Today, many modern routers are capable of ingress filtering. Many network administrators have turned on uRPF on their routers or been actively maintaining IALs to filter spoofing traffic. The fraction of networks where spoofing is possible is significantly limited [MIT-Spoofer]. However, as shown by the recent measurement, the deployment of ingress filtering has not been improved over four years. Sophisticated attackers can still exploit the networks where spoofing is possible to launch spoofing based attacks, and IP spoofing remains a viable attack vector on the current Internet [Efficacy]. In this section, we evaluate ingress filtering against the desired features, and analyze the reasons why ingress filtering is not sufficiently deployed, and why its deployment has not been improved over years. We classify the five variants of ingress filtering (IAL and four variants of uRPF) into three categories according to their technical similarities, i.e., IALs, strict/feasible RPF, and loose Bi & Liu Expires November 10, 2017 [Page 4] Internet-Draft SAVI Problem Beyond First Hop May 2017 RPF* (including loose RPF and loose RPF ignoring default routes). The evaluation results are summarized in Table 1, which will be explained in detail in the following subsections. +---------------------+-----------+------+------+ | Techniques | Incentive | Risk | Cost | +---------------------+-----------+------+------+ | IAL | low | low | low | | - | - | - | - | | Strict/feasible RPF | high | high | low | | - | - | - | - | | Loose RPF* | low | low | low | +---------------------+-----------+------+------+ Table 1: Evaluation of Ingress Filtering 3.1. IAL IAL can be applied at network border routers and enforced on outbound packets. All the outbound packets whose source addresses do not belong to the local network are identifed as spoofed. Since it only filters outbound packets, and cannot protect the local network from receiving spoofing packets, it provides little deployment incentives. When IAL is enforced on inbound packets, there are two typical scenarios. In the first scenario, it is applied on the routers' ingress interface connecting a stub network. In this case, the source address space of the stub network can be configured on this interface, and all inbound packets whose source addresses fall out of this address space are identified as spoofed. Since the address space size of directly connected stub networks is very small compared to the entire routable address space, the effectiveness is low. In the second scenario, the directly connected network is not stub, and thus the valid address space of inbound packets is hard to determine. In this case, IAL can only safely drop the inbound packets whose source addresses belong to the local network. It is very ineffective since the address space size of the local network very small compared to the entire routable address space. To summize, IAL provides low deployment incentives in all cases. If well maintained, false positives can also be avoided. IAL can be implemented using the existing functions on routers (access control lists, ACLs), and thus do not require additional investment. Overall, the deployment cost is low except in the "stub-network" scenario, where the address space of all stubs need to be carefully maintained, which may require heavy manual configuration if there are many directly connected stub networks. 3.2. Strict/feasible RPF Bi & Liu Expires November 10, 2017 [Page 5] Internet-Draft SAVI Problem Beyond First Hop May 2017 Strict and feasible RPF can adapt themselves to the routing dynamics and determine the incoming directions of source prefixes automatically. They are more effective than IAL in detecting and filtering spoofing traffic, and thus provide higher deployment incentives to the ISPs. However, they often drop legitimate packets under routing asymmetry, which is very prevalent with the existence of local routing policies, multi-homing and traffic engineering. This makes strict and feasible very risky [NANOG-Risk], and hence the operation needs to be very careful. Feasible RPF has lower FP than strict RPF, since it can apply multiple interfaces as acceptable incoming directions for a source prefix. But feasible RPF cannot avoid all FP in practice, since currently there is no practical way to generate and configure all possible incoming directions in the routers. For example, in a link-state routing environment (IS-IS or OSPF), equal-cost multi-path (ECMP) [ECMP] is often used to generate the multiple acceptable incoming directions. However, in practice, there can be many (tens of) ECMPs for a prefix, but the implementation of a router can only store several (e.g., 4 or 8) of them. Thus, the ECMPs for the prefix installed in the forwarding table may be different in different routers, which eventually causes the FP of feasible RPF. And in BGP, the directions where BGP announcements for a source address prefix have been received can be considered as acceptable incoming directions [IDPF]. However, an ISP may choose not to announce a prefix via a path but still send traffic through it due to its local routing policy. In this case, feasible RPF also causes FP. The basic function of strict and feasible RPF is supported in most modern routers. So deploying them doesn't require investment on upgrading devices. 3.3. Loose RPF* Instead of validating the incoming direction of a source address, loose RPF* only checks the existence of the source address in the forwarding table. Loose RPF* is only useful for filtering Martian addresses and unroutable addresses. However, sophisticated attackers can evade loose RPF* checking by simply using routable source addresses. Thus the incentive to deploy loose RPF is low. On the other hand, its operational risk is also low. Loose RPF* is also available in most modern routers, making it cheap to deploy. 3.4. Other Reasons There are other reasons why the deployment of ingress filtering hasn't been improved in four years. First, although most modern routers are capable of ingress filtering, some legacy routers are incapable [NANOG-Equipment]. Hence, even at the locations where no risk exists (e.g. stub or single-homed networks), ingress filtering may not be applied. Another reason is called inertia. Since ingress Bi & Liu Expires November 10, 2017 [Page 6] Internet-Draft SAVI Problem Beyond First Hop May 2017 filtering is not enabled on routers by default, some network administrators just won't bother to turn it on [NANOG-PowerOfDefaults]. 4. Discussion In this section, we discuss the directions that we can explore to improve SAVI-BF. We briefly survey two categories of SAVI-BF proposals, path based techniques and end-to-end based techniques. We discuss their requirements, advantages and disadvantages. We only focus on the techniques that are implemented on the routers. The techniques implemented on end hosts, such as [IPSec], [HCF] and [IP-Puzzles], are not covered here. 4.1. Path based Techniques The path based techniques essentially require that a router R knows the forwarding paths that each source prefix S uses toward its destinations, and subsequently knows the incoming directions/ interfaces of S. Sometimes, this information is available to R. For example, in a pure link-state routing protocol environment (e.g., IS- IS, OSPF), all nodes have the same view of the network. Thus, R can compute the paths from S to all destinations, and then infer the incoming directions of S. One exception is that, when there are multiple equally best paths, R may not determine which one S will use. On the other hand, if a distance-vector (e.g., RIP) or a path- vector (BGP) routing protocol is used, it is even harder for R to determine the paths of S, since the sufficient routing information is missed. [SAVE] and [IDPF] are tow proposals which validate source addresses by inferring their forwarding paths. 4.2. End-to-end based Techniques There are, however, other proposals that don't rely on path information. We call them end-to-end based techniques. For example, [SPM] associates each source prefix (indeed, source AS number) with a key. S, an SPM-enabled AS, will tag the key into outbound packets toward R at its border routers. And R verifies the keys of the inbound packets whose source addresses belong to S at its border routers. The routers will drop a packet if the key is incorrect. Thus, R manages to validate the source addresses of S without knowing its forwarding paths. The end-to-end based the techniques typically need the cooperation between the source and the verification nodes, and require particular tags be carried in the data packets. Further more, even the end-to-end based techniques require to distinguish inbound traffic and outbound traffic, which is not completely path- independent. Bi & Liu Expires November 10, 2017 [Page 7] Internet-Draft SAVI Problem Beyond First Hop May 2017 4.3. Non-technical Proposals There are also proposals which formulate source address validation as an economic problem [FaaS], or suggest that laws and governance should be enforced. These directions, however, may be out of the scope of IETF. 5. Acknowledgment The authors would like to thank Fred Baker and Joel M. Halpern for their comments. This document was generated using the xml2rfc tool. 6. Informative References [ARBOR-2010] Dobbins, R. and C. Morales, "Network Infrastructure Security Report", February 2010. [BCP38] Paul, P. and D. Senie, "Network Ingress Filtering: Defeating Denial of Service Attacks which employ IP Source Address Spoofing", RFC 2827, BCP 38, May 2000. [BCP84] Baker, F. and P. Savola, "Ingress Filtering for Multihomed Networks", RFC 3704, BCP 84, March 2004. [DrDoS-300Gbps] Prince, M., "The DDoS That Almost Broke the Internet", March 2013. [ECMP] Thaler, D. and C. Hopps, "Multipath Issues in Unicast and Multicast Next-Hop Selection", RFC 2991, November 2000. [Efficacy] Beverly, R., Berger, A., Hyun, Y., and k. claffy, "Understanding the Efficacy of Deployed Internet Source Address Validation Filtering", August 2009. [FaaS] Liu, B., Bi, J., and X. Yang, "FaaS: Filtering IP Spoofing Traffic as a Service", 2012. [Ground-Truth] Labovitz, C., "Botnets, DDoS and Ground-Truth", October 2010. [HCF] Jin, C., Wang, H., and K. Shin, "Hop-count filtering: an effective defense against spoofed DDoS traffic", 2003. Bi & Liu Expires November 10, 2017 [Page 8] Internet-Draft SAVI Problem Beyond First Hop May 2017 [IDPF] Duan, Z., Yuan, X., and J. Ch, "Controlling IP Spoofing Through Inter-Domain Packet Filters", 2008. [IP-Puzzles] Feng, W., Feng, W., and A. Luu, "The Design and Implementation of Network Puzzles", 2005. [IPSec] Kent, S. and K. Seo, "Security Architecture for the Internet Protocol", RFC 4301, December 2005. [MIT-Spoofer] Beverly, R., "Spoofer Project", January 2012, . [NANOG-Equipment] Bicknell, L., "BCP38 Deployment", March 2012, . [NANOG-Helpless] Bulk, F., "Are we really this helpless? (Re: isprime DOS in progress)", January 2009, . [NANOG-PowerOfDefaults] Donelan, S., "BCP38 Deployment", March 2012, . [NANOG-Risk] Gilmore, P., "BCP38 Deployment", March 2012, . [SAVE] Li, J., Mirkovic, J., Wang, M., Reiher, P., and L. Zhang, "SAVE: Source Address Validity Enforcement Protocol", 2002. [SPM] Anat, A. and H. Hanoch, "Spoofing Prevention Method", March 2005. Authors' Addresses Jun Bi Tsinghua University Network Research Center, Tsinghua University Beijing 100084 China Email: junbi@tsinghua.edu.cn Bi & Liu Expires November 10, 2017 [Page 9] Internet-Draft SAVI Problem Beyond First Hop May 2017 Bingyang Liu Tsinghua University Network Research Center, Tsinghua University Beijing 100084 China Email: bingyangliu.cn@gmail.com Bi & Liu Expires November 10, 2017 [Page 10]