Network Working Group R. Hibbs
Internet-DraftINTERNET-DRAFT Richard Barr Hibbs, P.E. Expires: October 29, 2004Category: Informational C. Smith Expires: December 15, 2006 C & C Catering B. Volz Cisco Systems, Inc. M. Zohar IBM T. J. Watson Research Center April 30, 2004June 13, 2006 Dynamic Host Configuration Protocol for IPv4 (DHCPv4) Threat Analysis draft-ietf-dhc-v4-threat-analysis-02 Status of this Memo<draft-ietf-dhc-v4-threat-analysis-03.txt> Saved: Tuesday, June 13, 2006, 12:56:25 Intellectual Property Rights By submitting this Internet-Draft, I certifyeach author represents that any applicable patent or other IPR claims of which I amhe or she is aware have been or will be disclosed, and any of which I becomehe or she becomes aware will be disclosed, in accordance with RFC 3668.Section 6 of BCP 79. Status of this Memo Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts.Internet- Drafts. 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." The list of current Internet-Drafts can be accessed at http:// www.ietf.org/ietf/1id-abstracts.txt.http://www.ietf.org/1id-abstracts.html. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on October 29, 2004.Comments are solicited and should be addressed to the working group's mailing list at firstname.lastname@example.org and/or the author(s). Copyright Notice Copyright (C) The Internet Society (2004). All Rights Reserved.(2006). Abstract DHCPv4 (RFC 2131) is a stable, widely used protocol for configuration of host systems in a TCP/IPv4 network. It did not provide for authentication of clients and servers, nor did it provide for data confidentiality. This is reflected in the original "Security Considerations" section of RFC 2131, which identifies a few threats and leaves development of any defenses against those threats to future work. Beginning inIn about 19951995, DHCP security began to attract attention from the Internet community, eventually resulting in the publication of RFC 3118 in 2001. Although RFC 3118 was a mandatory prerequisite for the DHCPv4 Reconfigure Extension, RFC 3203, it has had no known usage by any commercial or private implementation since its adoption. The DHC Working Group hasadopted a work item for 2003 to review and modify or replace RFC 3118 to afford a workable, easily deployed security mechanism for DHCPv4. This memo provides a comprehensivethreat analysis of the Dynamic Host Configuration Protocol for Ipv4 (DHCPv4) for use both as RFC 2131 advances from Draft Standard to Full Standard and to support our chartered work improving the acceptance and deployment of RFC 3118. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 41 Introduction................................................4 1.1 Issues for Consideration . . . . . . . . . . . . . . . . . 4Consideration...............................4 1.2 Assumptions . . . . . . . . . . . . . . . . . . . . . . . 4 1.3Exclusions.............................................4 2 Use of Key Words............................................5 3 Applicability...............................................5 3.1 Assumptions............................................5 3.2 Scope of this Memo . . . . . . . . . . . . . . . . . . . .Memo.....................................5 4 1.4 Use of Key Words . . . . . . . . . . . . . . . . . . . . . 5 2.General Threatsthreats to DHCPv4 . . . . . . . . . . . . . . . . . . 5 2.1DHCPv4...................................5 4.1 Denial-of-Service Attacks . . . . . . . . . . . . . . . . 5 2.1.1Attacks..............................5 4.1.1 Refusal to Configure Clients . . . . . . . . . . . . . 5 2.1.2Clients.....,...............5 4.1.2 Impersonating Clients . . . . . . . . . . . . . . . . 5 2.1.3 Flooding . . . . . . . . . . . . . . . . . . . . . . . 6 2.2Clients.............,..............5 4.1.3 Flooding...........................,.............6 4.2 Client Misconfiguration . . . . . . . . . . . . . . . . . 6 2.3Misconfiguration................................6 4.3 Theft of Network Service . . . . . . . . . . . . . . . . . 6 2.4Service...............................6 4.4 Packet Insertion, Deletion, or Modification . . . . . . . 7 3.Modification............7 5 Weaknesses of RFC 3118 . . . . . . . . . . . . . . . . . . . . 7 3.13118......................................7 5.1 Key Exposure . . . . . . . . . . . . . . . . . . . . . . . 7 3.2Exposure...........................................7 5.2 Key Distribution . . . . . . . . . . . . . . . . . . . . . 7 3.3Distribution.......................................7 5.3 Replay attacks . . . . . . . . . . . . . . . . . . . . . . 8 3.4attacks.........................................8 5.4 Protocol Agreement Difficulties . . . . . . . . . . . . . 8 4.Difficulties........................8 5.5 DHCPv4 Relay Agents Excluded...........................8 5.6 Unanticipated Infrastructure Changes...................8 6 DHCPv4 Security Requirements . . . . . . . . . . . . . . . . . 8 4.1 Environments . . . . . . . . . . . . . . . . . . . . . . . 8 4.2 Capabilities . . . . . . . . . . . . . . . . . . . . . . . 8 4.3Requirements................................9 6.1 Environments...........................................9 6.2 Capabilities..........................................10 6.3 Musings on the Key Distribution Problem . . . . . . . . . 9 4.4Problem...............10 6.4 Data Confidentiality . . . . . . . . . . . . . . . . . . . 10 4.4.1Confidentiality..................................11 6.4.1 "Public" Data in DHCP Packets . . . . . . . . . . . . 10 4.4.2Packets...................12 6.4.2 Protecting Data in DHCP Options . . . . . . . . . . . 11 5.Options.................12 6.5 Host versus User Authentication.......................12 6.5.1 Why do we make this distinction?................13 6.5.2 Is one mechanism sufficient?....................13 7 IANA Considerations . . . . . . . . . . . . . . . . . . . . . 11 6.Considerations........................................14 8 Security Considerations . . . . . . . . . . . . . . . . . . . 11 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 11 8. Change Log . . . . . . . . . . . . . . . . . . . . . . . . . . 11 8.1 01 Draft . . . . . . . . . . . . . . . . . . . . . . . . . 11 8.2 02 Draft . . . . . . . . . . . . . . . . . . . . . . . . . 12 9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 12 9.1Considerations....................................14 9 Acknowledgements...........................................14 10 References................................................14 10.1 Normative References . . . . . . . . . . . . . . . . . . . . 12 9.2References.................................14 10.2 Informative References . . . . . . . . . . . . . . . . . . . 13 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 14 Intellectual Property and Copyright Statements . . . . . . . . 15 1.References...............................15 1 Introduction DHCPv4 as defined in RFC 1541[RFC1541] and RFC 2131[RFC2131] does not provide any form of communication security, confidentiality, data integrity, or peer entity authentication. A design team was formed at IETF-55 in Atlanta in November 2002 to look at DHCPv4 and RFC 3118[RFC3118] to document security requirements for DHCPv4. RFC 3118[RFC3118] defines the current security mechanisms for DHCPv4. Unfortunately, RFC 3118 has neither been implemented nor deployed to date. There is widespread feeling that its current restriction to manual keying of clients limits its deployment. The DHC Working Group seeks to rectify this situation by defining security mechanisms for DHCPv4 that have better deployment properties. 1.1 Issues for Consideration Specific issues to be considered include: oO Improved key management and scalability. oO Security for messages passed between relay agents and servers. oO The increased usage of DHCPv4 on insecure (e.g., wireless) and public LANs. oO The need for clients to be able to authenticate servers, without simultaneously requiring client authentication by the server. 1.2 Assumptions This document assumes that the reader is familiar with bothO Does use of the base DHCPv4 protocol as defined in RFC 2131 andRelay Agent Information Option imply the DHCPv4 authentication extension as defined in RFC 3118, and does not attempt to provide a tutorial on either. 1.3 Scope of this Memo This document confines its analysis to DHCPv4, as defined in RFC 2131 and RFC 2132 [RFC2132] andneed for authenticated messages between DHCP Authentication, as defined in RFC 3118.servers and relay agents? 1.2 Exclusions Excluded from our analysis are: oO Securing messages between relay agents and servers: work is already underway on this, see [auth-subopt][RFC4030] and [relay-ipsec]. oO DHCP Reconfigure Extension (FORCERENEW), RFC 3203(FORCERENEW) [RFC3203]: the authors believe it is appropriate to put the onus to provide the analysis on those who are interested in moving that work forward. o DHCP Failover Protocol, as[Editor's note: despite repeated calls on the DHC Working Group mailing list to identify even a single implementation of FORCERENEW, we are unable to put forward an example of its use.] O DHCP Failover Protocol, as defined in [failover]: the server-to-serverserver- to-server protocol used differs significantly from DHCP. oDHCP, and there has been no recent work on the [failover] draft. O DHCP-DNS Interaction, as defined in [fqdn]: securing communication between DHCP servers and DNS servers is a DNS update security issue and therefore out of scope for the DHC working group. oO DHCPv6, as defined in RFC 3315[RFC3315]: while we believe that authentication techniques developed for DHCPv4 would generally be applicable to DHCPv6, there are fundamental differences between the two protocols and RFC 3118 specifies DHCPv4-style message and options formats, so we have chosen to concentrate on DHCPv4. oO DHCP Lease Query, as defined in [leasequery]:[RFC4388]: because of lack of maturity. 1.42 Use of Key Words The key words "MUST," "MUST NOT," "REQUIRED," "SHALL," "SHALL NOT," "SHOULD," "SHOULD NOT," "RECOMMENDED," "MAY," and "OPTIONAL" in this document are to be interpreted as described in RFC 2119[RFC2119]. 2.3 Applicability 3.1 Assumptions This document assumes that the reader is familiar with both the base DHCPv4 protocol as defined in [RFC2131] and the DHCPv4 authentication extension as defined in [RFC3118], and does not attempt to provide a tutorial on either. 3.2 Scope of this Memo This document confines its analysis to DHCPv4, as defined in [RFC2131] and [RFC2132] and DHCP Authentication, as defined in [RFC3118]. 4 General Threatsthreats to DHCPv4 These are the classes of threats to the base DHCPv4 protocol. Not all of these are DHCP-specific, nor can all the concerns listed be addressed by DHCP authentication. 2.14.1 Denial-of-Service Attacks 188.8.131.52.1 Refusal to Configure Clients A rogue DHCP server can refuse to configure clients by responding with either partial information (i.e., missing the IP address, yet containing other information)information), or a non-routable (or otherwise bad) IP address. Or,address, or the server may respond to DHCPDISCOVER messages (with DHCPOFFER messages) but then ignore the subsequent client DHCPREQUEST messages. This may cause a client to repeatedly fail to be configured, though clients could take steps to ensure that they subsequently ignore such servers for a period ofsome time. 184.108.40.206.2 Impersonating Clients A rogue client can impersonate a client or many clients, by using another client's client identiferidentifier (client identifier option) and/or hardware address (chaddr) or by generating these identifiers. This may be done to: oO Obtain addresses when hardware address or client identifier restrictions (lists) are configured into the site's server through some mechanism not described in RFC 2131. Some sites may use such a mechanism to restrict the clients that are allowed addresses. A rogue client listens to DHCPv4 traffic and captures a few chaddrs or client identifiers and starts using them. oO Simulate many clients to consume all available addresses. The rogue client may either hold on to these addresses (until the leases expire) or decline the addresses (by sending a DHCPDECLINE) in the hopes that the server will remove the declined address from use for a longer period of time. operiod. O Create havoc on the subnet by injecting fake messages on behalf of other clients thatclients, prematurely releasereleasing (DHCPRELEASE) or declinedeclining (DHCPDECLINE) their addresses. A rogue client listens to DHCPv4 traffic and gleansgleams client identity and address information and uses this information to inject fake messages. 220.127.116.11.3 Flooding A rogue client can flood the network with (near-) continuous DHCPv4 request messages thereby consuming processing resources and network bandwidth. We mention this attack only for completeness, as there is little or nothing that a DHCP server can do to prevent such an attack and the client could just as well send messages of other protocols; andprotocols, so we will not discuss it further. 2.24.2 Client Misconfiguration Rogue servers may give out bad configuration information (such as fake gateways or DNS servers), or relay agents or other network elements may alter packets between a client and server, to cause the client to be misconfigured, or to enablepotentially worse cause future man-in-the-middleman- in-the-middle attacks. This category is usually part of another attack, be it theft of service, business espionage, or business interruption including denial of service. 2.34.3 Theft of Network Service By "theft of network service"service", we mean the taking of an unused address for network access or the use of an assigned address not belonging to the client, in contrast with "client masquerading" (Section 2.1.2) which refers specifically to the use of a legitimate client's chaddr or client identifier. Instantiation of an unauthorized client for purposes of using network resources or services is only partially preventable using client-server authentication techniques. We mention this attack only for completeness, as there is little or nothing thata DHCP server, itself,server itself can do to prevent such an attack. Additional host and application security is required to prevent theft of service, and such layer 5 and higher functions are declared out of scope for this analysis. 2.44.4 Packet Insertion, Deletion, or Modification If a client (or server or relay agent) is known to crash or shut down when invalid packets of some type are sent, this could be simply another type of denial of service attack. Likewise, simply deleting certain packet types (DHCPREQUEST to renew or rebind a lease) would eventually result in client lease expiration, a denial of service attack. A rogue relay agent or other host would typically use packet insertion and deletion to interrupt service. In a different vein, the modification of packets in the DHCP exchange may be used to facilitate many different types of attacks on either client or server. For example, a DHCPACK could be modified to a DHCPNAK, thereby denying the client a lease. 3.5 Weaknesses of RFC 3118 An authentication mechanism for DHCPv4 protocol messages was developed in RFC 3118, proposing two basic authentication mechanisms and the means for extending the repertoire of methods as needed. The configuration token method (protocol 0) relies on exchanging clear-text authentication tokens between as yetunconfigured DHCPv4 clients and DHCPv4 servers. It is also vulnerable to message interception. Delayed authentication (protocol 1) focuses on solving the intradomain authentication problem where the out-of-band exchange of a shared secret is feasible. 3.15.1 Key Exposure The configuration token protocol, protocol 0, utilizes clear-text authentication tokens (i.e., passwords), providing only weak entity authentication and no message authentication. This protocol is vulnerable to interception and provides only the most rudimentary protection against inadvertently instantiated DHCP servers. It also leaks the key before knowing whether the server supports protocol 0. 3.25.2 Key Distribution Both protocols 0 and 1 suffer from the lack of a means to easily, quickly, and reliably distribute authentication tokens used in the protocols. In many environments, some existing key distribution mechanism is presumed to be trusted and reliable, with strong administrative procedures and a security-conscious user population in place, leaving only the selection and specification of an appropriate cryptographic algorithm as a concern of the protocol designer. Relying on such out-of-band methods to distribute and manage tens or hundreds of thousands of tokens is a significant barrier to the widespread implementation of either protocol 0 or 1. Key distribution presents a significant system management challenge that is in marked contrast with DHCP itself, a protocol that has been successfully deployed in environments that make few demands or assumptions. If we are to hope for similarly successful deployment of authentication for DHCP, a means for mitigating (if not eliminating) these difficulties must be offered. 3.35.3 Replay attacks Since the configuration token protocol, protocol 0, passes a clear-textclear- text authentication token, the token would be visible to any host on the same subnet. Delayed authentication, protocol 1, is not susceptible to replay attacks since it contains a nonce value generated by the source and a message authentication code (MAC) which provides both message and entity authentication. 3.45.4 Protocol Agreement Difficulties An a priori agreement is presumed to have taken place between client and server on the authentication protocol to use. No mechanism is provided to allow for the discovery of supported protocols, nor is there a facility for negotiation. The only way to express non-supportnon- support of a protocol is by failing to respond. 4. DHCPv4 Security Requirements5.5 DHCPv4 was developed in an era when computers were primarily used in business and university environments. Security was either notRelay Agents Excluded [RFC3118] is defined exclusively for client-server communication. The role of relay agents has expanded somewhat from their earliest definition to include a concern or was addressedDHCP option carrying relay agent information via sub-options [RFC3046]. An authentication sub-option for the relay agent information option has been defined by controlling physical access stemming from[RFC4030], though it only defines a single protocol, symmetrical shared-secret keys, to protect the belief that intrusion into critical systems was most likelycontents of the messages between DHCP relay agents and servers. Work-in-progress to come from an external source. Now, with wireless access pointsprotect the interaction between relay agents and ubiquitous networking, physical access control mechanisms areservers using IPSEC [relay-ipsec] seems to have halted, with no longer sufficient,recent work. 5.6 Unanticipated Infrastructure Changes Rapid commit, defined by [RFC4039], specifies how a two-message exchange between client and securityserver can dramatically decrease the elapsed time for address assignment, a feature becoming significant as more and privacy issues aremore highly mobile devices desire an abbreviated address assignment phase for short duration communications. While a major concern. 4.1 Environments The following environments cantwo-message exchange by itself does not affect the overall security of the communications, it has two side effects. First, the delayed authentication protocol simply cannot be expected for DHCPv4 implementations: o Network size fromused as the DHCPOFFER message required to return a few hostsnonce value to hundredsthe client is not present. Second, as noted by the authors of thousands[RFC4039], without authentication it is considerably more likely to consume addresses, increasing the risk of hosts. o Network topology from a single subnetone type of denial of service attack. CableLabs client configuration [RFC3495] adds another two-tier option to Class-A networks. o Network location from a single roomthe DHCPv4 options list, defining an initial set of sub- options for passing configuration information specific to international dispersion. o Wired, broadcast wireless,CableLabs VoIP and directional wireless media. o Movement between different mediapossible future services. Although several of the sub- options contain potentially sensitive information regarding Kerberos tickets to be used by cable modems and networks. 4.2 Capabilitiesmedia terminal adapters. The following are essential elementsoption specification does not require used of DHCPv4 security: o Clients MUSTDHCP authentication, although it would seem to be ablenecessary. The authors of this memo do not wish to authenticate servers (to prevent misconfigured clients and assurego so far as to recommend that the correct servers are being contacted). o Servers MUSTDHCP authentication be able to authenticate clients (use of hardware addresses and client-IDs provides no real security but isa requirement for any or all DHCP options, but the CableLabs client configuration option will likely not be the last option that could benefit from a robust, workable authentication implementation. Passive duplicate address detection [p-dad] is easily possible today). Better mechanisms are needed for servers to identify clientsa relatively new proposal to which they will offer service (to prevent IPreplace the use of ICMPECHO and ARP messages by DHCP clients and servers to improve the duplicate address pool depletion, for example). o Administrators MUST be abledetection process by passively listening to choose between four authentication paradigms: * No authentication required. * Mutual authentication required. * Client authenticates server. * Server authenticates client. o Integritymessage traffic and developing a table of matches between chaddr, DHCP packet exchanges MUST be assured. Not all capabilities mayclient identifier, and IP address that would be needed or desired in all situations. 4.3 Musings on the Key Distribution Problemperiodically transmitted to interested DHCP servers. The authors believepresence of an entry for a particular IP address would signify that only by addressing scalability issues with key distribution can RFC 3118 achieve wide deployment. Whileit is not our intentionknown to describe solutionsbe in this document, we admit that we finduse, so a DHCP server could exclude the model used todayaddress from its pool of addresses available for assignment. The address usage collector (AUC) defined by browsers and secure web servers attractive: trusted root certificates are distributed with[p-dad] does not use the clientDHCP client-server protocol, nor does it function by actively handling DHCP messages, so its implementation (web browser); users havewould not affect authenticated DHCP messages. However, DHC Working Group discussion of the ability to extend[p-dad] draft raised the certificatespoint that they will accept, install their own certificates (shouldthe AUC must capture the client identificationidentifiers used in DHCP message exchanges. If DHCP authentication were enabled, an AUC of necessity would be required), and choose which certificate to presentrequired to servers requestinghave the client's identity. Analogously, DHCPv4 servers could make use of certificates justsame authentication configuration data (protocol, algorithm, and key) as web servers do, while DHCPv4the clients could be distributed with appropriate certificate authority certificates (trust anchors). Self-signed certificates are, of course, a possibility, should an organization wish full control over its domain of trust. Should this path be pursued, we believe that certificate revocation will beand servers, certainly a major problem to confront, just as it is in the browser/ web server environment today. Revocation of client certificates (which we believe would occur, on the whole, much more frequently than revocation of server certificates), would require only ordinary careconsideration for scalability and risk assessment. 6 DHCPv4 Security Requirements DHCPv4 was developed in certificate validation by the DHCP server. Revocation of server certificates is more complex because of the difficulty updating client configurations, as well as the inability of clients to rely on certificate revocation lists whilean era when computers were primarily used in the process of performing IP addressbusiness and configuration management. 4.4 Data Confidentiality Data Confidentialityuniversity environments. Security was either not provided for in the original DHCP protocol as defined in RFC 2131a concern or any ofwas addressed by controlling physical access stemming from the subsequent RFCs. Historically, DHCPbelief that intrusion into critical systems was mainly usedmost likely to assign IP addresses and return configuration options such as the local gatewaycome from an external source. Now, with wireless access points and DNS information. Over time the DHCP protocol has evolved, deploymentsubiquitous networking, physical access control mechanisms are extending beyond physically secure intranets to public networks in hotspots, cafes, airports,no longer sufficient, and at home over broadband. And wesecurity and privacy issues are seeing an accompanying proliferation of new configuration options. DHCP has, in fact, become so successful that it is now used to transporta great deal of configuration data that couldmajor concern. 6.1 Environments The following environments can be obtained inexpected for DHCPv4 implementations: O Network size, from a varietyfew hosts to hundreds of other ways. It is certainly possible thatthousands of hosts. O Network topology, from a client or server might wishsingle subnet to reveal someClass-A networks. O Network location, from a single room to international dispersion. O Wired, broadcast wireless, and directional wireless media. O Movement between different media and networks. 6.2 Capabilities The following are essential elements of these data onlyDHCPv4 security: 1. Clients MUST be able to a properly-authenticated peer. 4.4.1 "Public" Data in DHCP Packets We assume that any informationauthenticate servers (to prevent misconfigured clients and assure that may be gleaned directly fromthe network using, for example, Ethernet promiscuous mode is not confidential. It couldcorrect servers are being contacted). 2. Servers MUST be argued that over time moreable to authenticate clients (use of hardware addresses and more communication will be switched, encrypted, or secured at the physical layer, soclient-IDs provides no real security but is all that less information couldis easily be gleaned from the network traffic. Taking encryption into consideration, thepossible today). Better mechanisms are needed for servers to identify clients to whom they will offer service (to prevent IP address pool depletion, for example). 3. Administrators MUST be able to choose between four authentication paradigms: a. No authentication required. b. Mutual authentication required. c. Client authenticates server. d. Server authenticates client. 4. Integrity of DHCP packet payload mightexchanges MUST be encrypted, but notassured. Not all capabilities may be needed or desired in all situations. 6.3 Musings on the IP header itself sinceKey Distribution Problem The authors believe that only by addressing scalability issues with key distribution can RFC 3118 achieve wide deployment. While it is required for packet routing. As a result none of the IP header fields are confidential. IP addresses included in the header are therefore not confidential. Similarly, the hardware addresses are also not confidential. Although the IP packet payload (which would include the UDP or TCP header) might normally be encrypted, some protocols have made explicit decisionsnot our intention to encrypt their exchanges, declaring their data public. DNS is such a protocol [dns-threats]. Thusdescribe solutions in this document, we may also treat DNS domainadmit that we find several models used today by browsers and server informationsecure web servers as public. Commonly-used routing protocolswell as token-based user authentication schemes such as BGP (RFC 1771) [RFC1771], RIP (RFC 1721) [RFC1721], and router discovery (RFC 1256) [RFC1256] also normally send advertisements inthe clear and we therefore extend our definition of public DHCP dataRSA SecureID token to include routing information options. 4.4.2 Protecting Data in DHCP Options Some DHCP options (e.g., relay agent options, RFC 3046 [RFC3046]) or option families (site or vendor options) admit no analysis because the data carried by them maybe of unknown sensitivity. Analysis ofattractive. Trusted root certificates are distributed with the client implementation (web browser); users have the ability to extend the certificates that they will accept, install their confidentiality requirements mustown certificates (should client identification be done by their users. Other DHCPv4 options contain opaque data, not subjectrequired), and choose which certificate to interpretation by a DHCPv4 server. With no RFC-based definition ofpresent to servers requesting the data content of these options, we can offer no opinion ofclient's identity. Security tokens that combine a secure seed value with the sensitivitycurrent time of the data, and leave any confidentiality treatment to other work. Should some data require confidentiality, it may be possible to exploit the "public" data aboveday using a cryptographic algorithm to allowproduce effectively a two-step configuration process in which sufficient client configuration is first obtained by the normal DHCPDISCOVER/OFFER/REQUEST/ACK exchange,random one-time pad are relatively inexpensive and private data subsequently transmitted over a secure communications channel (suchwidely available. Analogously, DHCPv4 servers could make use of certificates just as IPsec) using DHCPINFORM. 5. IANA Considerations None known. 6. Security Considerations This entire memo presentsweb servers do, while DHCPv4 clients could be distributed with appropriate certificate authority certificates (trust anchors). Self-signed certificates are, of course, a threat analysispossibility should an organization wish full control over its domain of the DHCPv4 protocol. 7. Acknowledgements This documenttrust. Should this path be pursued, we believe that certificate revocation will be a major problem to confront, just as it is in the resultbrowser/web server environment today. Revocation of work undertakenclient certificates (which we believe would occur, on the whole, much more frequently than revocation of server certificates) would require only ordinary care in certificate validation by the DHCP working group, beginning at 55th IETF meeting in Atlanta. The authors would also likeserver. Revocation of server certificates is more complex because of the difficulty updating client configurations, as well as the inability of clients to acknowledge contributions from others not directly involvedrely on certificate revocation lists while in writing this memo, including John Beatty and Vipul Guptathe process of Sun Microsystems,performing IP address and Ralph Dromsconfiguration management. Using a security token device today is mostly to identify a person requesting access to a private resource. Key fobs, USB dongles, or wallet cards are in the possession of Cisco Systems. And Bernard Abobaa user, and Mark Stapp for their careful reviewsin conjunction with a user name, password, and suggestions. 8. Change Log This section summariespossibly other information confirms two of the changes made tothree classic dimensions of provable identity (something you know--user name and password, something you have--the security token, and something you are--biometrics typically satisfy this Internet-Draft as it has evolved. 8.1 01 Draft No significant changes were made: o Updated author information. o Converteddimension.) We envision a security token becoming part of the host system's hardware complement in the near future, such that the token then becomes not a user identity validator, but a host system validator. It is common today to using xml2rfchave a system service tag or serial number that is machine-readable. Some hardware configurations include processors with readable serial numbers as well. What we lack is a secure means to generate a cryptographically random key that cannot be easily defeated by software or component swapping. Either of these approaches offers a simple way to avoid the draft. o Removed unused references. o Addedclassic key distribution problem, though neither is totally without cost. Somehow, somewhere, a token's identifiers (seed and algorithm) must be recorded by the Change Log section. 8.2 02 Draft The following changes were made: o Updated author information. o Added text to 1.3 to exclude security for messages passed between relay agentsDHCP and servers as there are two Internet-Drafts on this subject. o Reworded several sections, particularilyother interested servers, or the certificate infrastructure must be in section 2. o Revised and renamed section 2.1.2; now includes more attacks. o Revised section 2.1.3. o Minor revisionsplace. We see no way to section 3, 3.2, and 3.2. o Added moreeliminate administrative issues associated with security, but we can see an end to section 4.4.2. o Other minor insertions, deletions, and modifications basedpasswords written on comments from Bernard Aboba and Mark Stapp and to otherwise improvea sticky note. An interesting Internet-Draft by Alper Yegin et al. [eap-auth] proposed the document. 9. References 9.1 Normative References [RFC1256] Deering, S., "ICMP Router Discovery Messages", RFC 1256, September 1991. [RFC1541] Droms, R., "Dynamic Host Configuration Protocol", RFC 1541, October 1993. [RFC1721] Malkin, G., "RIP Version 2 Protocol Analysis", RFC 1721, November 1994. [RFC1771] Rekhter, Y.use of EAP for DHCP authentication using the delayed method. Their draft required modifications to several components of an AAA solution, but illustrated how delayed authentication could be "bootstrapped" using tools at our disposal. That idea is also suggested by [RFC4014] Ralph Droms and T. Li, "A Border Gateway Protocol 4 (BGP-4)", RFC 1771, March 1995. [RFC2119] Bradner, S., "Key wordsJohn Schnizlein, who define a DHCP option code for useRADIUS information provided by an NAS, similar to the mechanism in RFCs[eap-auth]. These two documents taken together may provide a third solution to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC2131] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131, March 1997. [RFC2132] Alexander, S. and R. Droms, "DHCP Options and BOOTP Vendor Extensions", RFC 2132, March 1997. [RFC3046] Patrick, M., "DHCP Relay Agent Information Option", RFC 3046, January 2001. [RFC3118] Droms, R. and W. Arbaugh, "Authenticationthe key distribution problem. 6.4 Data Confidentiality Data Confidentiality was not provided for in the original DHCP Messages", RFC 3118, June 2001. 9.2 Informative References [RFC3203] T'Joens, Y., Hublet, C. and P. De Schrijver, "DHCP reconfigure extension",protocol as defined in RFC 3203, December 2001. [RFC3315] Droms, R., Bound, J., Volz, B., Lemon, T., Perkins, C.2131 or any of the subsequent RFCs. Historically, DHCP was mainly used to assign IP addresses and M. Carney, "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)", RFC 3315, July 2003. [auth-subopt] Stapp, M.return configuration options such as the local gateway and T. Lemon, "The Authentication Suboption forDNS information. Over time the DHCP Relay Agent Option", draft-ietf-dhc-auth-suboption-03 (workprotocol has evolved, deployments are extending beyond physically secure intranets to public networks in progress),hotspots, cafes, airports, and at home over broadband. We are seeing an accompanying proliferation of new configuration options. DHCP has, in fact, become so successful that it is now used to transport a great deal of configuration data that could be obtained in a variety of other ways. It is certainly possible that a client or server will wish to reveal some of these data only to a properly authenticated peer. 6.4.1 "Public" Data in DHCP Packets We assume that any information that may be gleaned directly from the network using, for example, Ethernet promiscuous mode is not confidential. It could be argued that over time more and more communication will be switched, encrypted, or secured at the physical layer, so that less information could easily be gleaned from the network traffic. Taking encryption into consideration, the IP packet payload might be encrypted, but not the IP header itself since it is required for packet routing. As a result, none of the IP header fields are confidential. IP addresses included in the header are therefore not confidential. Similarly, the hardware addresses are also not confidential. Although the IP packet payload (which would include the UDP or TCP header) might normally be encrypted, some protocols have made explicit decisions not to encrypt their exchanges, declaring their data public. DNS is such a protocol [dns-threats]. Thus, we may also treat DNS domain and server information as public. Commonly used routing protocols such as BGP [RFC1771], RIP [RFC1721], and router discovery [RFC1256] also normally send advertisements in the clear and we therefore extend our treatment of public DHCP data to routing information. 6.4.2 Protecting Data in DHCP Options Some DHCP options (e.g., relay agent options, [RFC3046]) or option families (site or vendor options) admit no analysis because the data carried by them may be of unknown sensitivity. Users must do their own analysis of confidentiality. Should some data require confidentiality, it may be possible to exploit the "public" data above to allow a two-step configuration process in which sufficient client configuration is first obtained by the normal DHCPDISCOVER/OFFER/REQUEST/ACK exchange, and private data subsequently transmitted over a secure communications channel (such as IPsec) using DHCPINFORM. 6.5 Host versus User Authentication [RFC3118] is concerned specifically with DHCP clients and servers authenticating themselves to each other if required by an administrative domain. This is not the same thing as authenticating users for establishing their Identity, access rights, permissions, or other matters relating to what they can view or do once connected to the network. 6.5.1 Why do we make this distinction? Host authentication provides only assurance that the hosts connecting to a network are recognized. This may be for several reasons, including: O Requirement to restrict network access from "foreign" hosts to ensure consistent technical support or meet other regulatory requirements such as double-insulation or non-sparking. O Requirement to restrict network access from "unsecured" (for instance, non-TEMPEST compliant) hosts in a high security network. O Requirement to restrict network access from unknown hosts whose identity has not been recorded by existing administrative procedures, saving troubleshooting and administrative effort. User authentication focuses on the individual users of a host system, seeking to uniquely identify someone to establish their rights to view, print, add, alter, delete, edit, modify, copy, or manipulate any data maintained by an organization, run software programs, or affect changes in the environment. This may be for such reasons as: O Requirement to be compliant with regulations such as HIPAA, SOX, and GLBA designed to safeguard and protect confidentiality and various reporting requirements. O Requirement to maintain reasonable controls over access to certain critical systems such as utility power grids, water and sewage treatment plants, the network infrastructure itself, and life safety systems of all sorts. O Requirement to maintain administrative controls over certain sensitive information such as trade secrets and personnel data 6.5.2 Is one mechanism sufficient? Full discussion of this question is beyond the scope of this memo, but the simple answer is, "probably not." This has a significant implication for the claims of certain techniques such as "sandboxing" of unverified users, restricting their network access to a user registration web site until their identity has been established. Simply put, limiting network access is not DHCP authentication, although it does represent a very workable approach to user authentication in many cases. Recall that a user can be identified by "something they know," "something they have," and "something they are." While the same could be said to be true of host systems, the authors point out that while we do not pretend to understand all of the ways that future developers might imbue hosts with the tools to independently create attacks on our infrastructure, we will assert that users will be the greatest risk for some time still. DHCP Authentication addresses only host authentication from the belief that other, existing mechanisms such as IPSEC, SSL, and VPN can protect the content of communications, with Identity Management and other technologies can protect applications and data. A properly authenticated host could still launch a denial of service attack, corrupt sensitive data, or wreak other havoc, which underscores our point that host and user authentication are different. A well-secured network needs both. 7 IANA Considerations None known. 8 Security Considerations This entire memo presents a threat analysis of the DHCPv4 protocol. 9 Acknowledgements This document is the result of work undertaken the by DHCP working group, beginning at the 55th IETF meeting in Atlanta. The authors would also like to acknowledge contributions from others not directly involved in writing this memo, including John Beatty and Vipul Gupta of Sun Microsystems, Ralph Droms of Cisco Systems, Bernard Aboba of Microsoft, and Mark Stapp of Cisco Systems for their careful reviews and helpful suggestions. 10 References 10.1 Normative References [RFC1256] Deering, S., "ICMP Router Discovery Messages," RFC 1256, September 1991. [RFC1541] Droms, R., "Dynamic Host Configuration Protocol," RFC 1541, October 1993. [RFC1721] Malkin, G., "RIP Version 2 Protocol Analysis," RFC 1721, November 1994. [RFC1771] Rekhter, Y. and T. Li, "A Border Gateway Protocol 4 (BGP- 4)," RFC 1771, March 1995. [RFC2131] Droms, R., "Dynamic Host Configuration Protocol," RFC 2131, March 1997. [RFC2132] Alexander, S. and R. Droms, "DHCP Options and BOOTP Vendor Extensions", RFC 2132, March 1997. [RFC3046] Patrick, M., "DHCP Relay Agent Information Option," RFC 3046, January 2001. [RFC3118] Droms, R. and W. Arbaugh, "Authentication for DHCP Messages," RFC 3118, June 2001. [RFC4014] Droms, R. and J. Schnizlein, " Remote Authentication Dial- In User Service (RADIUS) Attributes Suboption for the Dynamic Host Configuration Protocol (DHCP) Relay Agent Information Option," RFC 4014, February 2005. 10.2 Informative References [RFC2026] Bradner, S., "The Internet Standards Process -- Revision 3," RFC 2026, BCP 9, October 1996. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC3203] T'Joens, Y., C. Hublet and P. De Schrijver, "DHCP Reconfigure Extension," RFC 3203, December 2001. [RFC3315] Droms, R., Bound, J., Volz, B., Lemon, T., Perkins, C. and M. Carney, "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)," RFC 3315, July 2003. [RFC3495] Beser, B. and P. Duffy, "Dynamic Host Configuration Protocol (DHCP) Option for CableLabs Client Configuration," RFC 3495, March 2003. [RFC3594] Duffy, P., PacketCable Security Ticket Control Sub-Option for the DHCP CableLabs Client Configuration (CCC) Option"," RFC 3594, September 2003. [RFC3978] Bradner, S., "IETF Rights in Contributions," RFC 3978, BCP 78, March 2005. [RFC3979] Bradner, S., "Intellectual Property Rights in IETF Technology," RFC 3979, BCP 79, March 2005. [RFC4030] Stapp, M. and T. Lemon, "The Authentication Suboption for the DHCP Relay Agent Option," draft-ietf-dhc-auth-suboption-03), RFC 4030, March 2005. [RFC4039] Park, S., P. Kim and B. Volz, "Rapid Commit Option for the Dynamic Host Configuration Protocol version 4 (DHCPv4)," RFC 4039, March 2005. [RFC4388] Woundy, R., and Kinnear, K., "Dynamic Host Configuration Protocol (DHCP) Leasequery," February 2004.2006. [dns-threats] Atkins, D. and R. Austein, "Threat Analysis Of The Domain Name System", draft-ietf-dnsext-dns-threats-06System," draft-ietf-dnsext-dns-threats-07 (work in progress), FebruaryApril 2004. [draft2223bis] Reynolds, J. and R. Braden, "Instructions to Request for Comments (RFC) Authors," draft-rfc-editor-rfc2223bis-08.txt (work in progress), August 2004. [eap-auth] Yegin, A., H. Tschofenig and D. Forsberg, "Bootstrapping RFC3118 Delayed DHCP Authentication Using EAP-based Network Access Authentication," draft-yegin-eap-boot-rfc3118-02, (work in progress), March 2006. [failover] Droms, R. and K. Kinnear, "DHCP Failover Protocol",Protocol," draft-ietf-dhc-failover-12 (work in progress), December 2003. [Editor's note: at the time of publication of this memo, the Failover Internet-Draft has expired.] [fqdn] Stapp, M.M., Volz, B., and Y. Rekhter, "The DHCP Client FQDN Option", draft-ietf-dhc-fqdn-option-06 (work in progress), October 2003. [leasequery] Woundy, R., "DHCP Lease Query", draft-ietf-dhc-leasequery-07Option," draft-ietf-dhc-fqdn-option-13 (work in progress), March 2004.2006. [p-dad] Forte, A., S. Shin and H. Schulzrinne, "Passive Duplicate Address Detection for Dynamic Host Configuration Protocol (DHCP)," draft-forte-dhc-passive-dad-02 (work-in-progress), June 2006. [relay-ipsec] Droms, R., "Authentication of DHCP Relay Agent Options Using IPsec", draft-ietf-dhc-relay-agent-ipsec-01Ipsec," draft-ietf-dhc-relay-agent-ipsec-02 (work in progress), November 2003.May 2005. [Editor's note: at the time of publication of this memo, the Relay IPSEC Internet-Draft has expired.] Authors' Addresses Richard Barr Hibbs Richard Barr Hibbs, P.E. 952 Sanchez Street San Francisco, California 94114-3362 USA Phone: +1 415 648 3920 Fax: +1 415 648 9017 EMail:E-Mail: email@example.com Carl Smith C & C Catering 1121 Holly StStreet Alameda, California 94502 USA EMail:E-Mail: firstname.lastname@example.org Bernard Volz Cisco Systems, Inc. 1414 Massachusetts Ave.Avenue Boxborough, MAMassachusetts 01719 USA Phone: +1 978 936 0382 EMail:E-Mail: email@example.com Mimi Zohar IBM T. J. Watson Research Center 19 Skyline Drive Hawthorne, New York 10532-2134 USA Phone: +1 914 784 7606 EMail:E-Mail: firstname.lastname@example.org Full Copyright Statement Copyright (C) The Internet Society (2006). All rights reserved. This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Intellectual Property StatementThe IETF takes no position regarding the validity or scopevalidity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf- email@example.com. Acknowledgement Funding for the RFC Editor function is provided by the IETF Administrative Support Activity (IASA). APPENDIX: NOTES This appendix will be removed in its entirety before the memo goes to Working Group Last Call. ISSUES LIST This section summarizes issues raised in this memo that require resolution by the DHC Working Group. 1. Because of the specific exception for inclusion of the Relay Agent Information Option [RFC3046] in cases where it does not fit in the main payload portion of a DHCP response packet, should the use of any Intellectual Property Rights or other rights that mightDHCP Authentication be claimed to pertainmandated in place of the Relay Agent Authentication suboption? 2. Does the CableLabs Client Configuration option [RFC3495] require the use of DHCP Authentication to protect sensitive information about Kerberos domains and keys? 3. Should the DHC Working Group promote the development of a new authentication protocol based on the implementation oruse of certificates? 4. Should the technology described in this document orDHC Working Group solicit an update from the extent to which any license under such rights might or might not be available; nor doesauthors of the EAP-based authentication protocol [eap-auth] and develop it represent thatas a new authentication protocol? 5. Should the DHC Working Group promote the development of a new authentication protocol based on hardware security tokens? CHANGE LOG This section summarizes the changes made to this memo as it has evolved. "-01" Draft No significant changes were made any independent effort to identify any such rights. Information onfrom initial ("-0") version: O Updated author information. O Removed unused references. O Added the IETF's procedures with respectChange Log section. "-02" Draft The following changes were made: O Updated author information. O Added text to rights in IETF Documents can be found1.3 to exclude security for messages passed between relay agents and servers, as there are two Internet- Drafts on this subject. O Reworded several sections in BCP 78section 2. O Revised and BCP 79. Copies of IPR disclosures maderenamed section 2.1.2. Now includes more attacks. O Revised section 2.1.3. O Minor revisions to the IETF Secretariatsection 3, 3.2, and 3.2. O Other minor insertions, deletions, and modifications based on comments from Bernard Aboba and Mark Stapp and any assurances of licensesto be made available, orotherwise improve the document. "-03" Draft The draft was updated, correcting minor spelling, grammatical, and typographical errors, and modified in the result of an attempt madefollowing ways: O Removed Section 8, "Change Log," to obtain a general license or permission forAPPENDIX and added an issues list section. O Replaced all Internet-Draft boilerplate with the usemost current versions. O Renumbered document sections. O Updated author information. O Updated references for I-Ds advanced to RFCs. O Added normative and informative references. O Added discussion of such proprietary rights by implementers or usersRelay Agents (section 3.5.) O Added section 3.6, discussing the side effects of this specification can be obtainedinfrastruture changes from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. PleaseRapid Commit and CableLabs configuration options, and the newly proposed Address Usage Collector (AUC) for passive duplicate address detection. O Expanded section 4.3, "Musings on the informationKey Distribution Problem" to the IETF at firstname.lastname@example.org. Disclaimerinclude description of Validity This document andhardware-based key generation tokens. O Added section 4.5, "Host versus User Authentication," to help clarify the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Copyright Statement Copyright (C) The Internet Society (2004). This documentproblem [RFC3118] is subjectintended to address. O Added discussion of the rights, licenses and restrictions containedRADIUS-based authentication proposal described in BCP 78, and except as set forth therein, the authors retain all their rights. Acknowledgment Funding for[RFC4014]. O Added discussion of the RFC Editor function is currently providedEAP-based authentication protocol proposed by the Internet Society.A. Yegin et al. O Inserted Intellectual Property Rights statement on first page. O Performed general spelling, grammar, and typography update of entire memo text. O Reviewed all drafts used as references, updated as necessary.