SPKI Requirements Carl M. Ellison INTERNET-DRAFT CyberCash, Inc. Expires: 22 September 97 17 March 1997 SPKI Requirements ---- ------------ Status of This Document This document is a compilation of the requirements for SPKI [Simple Public Key Infrastructure] certificates gathered from the discussion on the SPKI Working Group mailing list. Distribution of this document is unlimited. Comments should be sent to the SPKI (Simple Public Key Infrastructure) Working Group mailing list or to the author. This document is an Internet-Draft. 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 are draft documents valid for a maximum of six months. Internet-Drafts may be updated, replaced, or obsoleted by other documents at any time. It is not appropriate to use Internet- Drafts as reference material or to cite them other than as a ``working draft'' or ``work in progress.'' To learn the current status of any Internet-Draft, please check the 1id-abstracts.txt listing contained in the Internet-Drafts Shadow Directories on ds.internic.net (East USA), ftp.isi.edu (West USA), nic.nordu.net (North Europe), ftp.nis.garr.it (South Europe), munnari.oz.au (Pacific Rim), or ftp.is.co.za (Africa). Ellison [Page 1] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 Abstract The IETF Simple Public Key Infrastructure [SPKI] Working Group is tasked with producing a certificate structure and operating procedure to meet the needs of the Internet community for trust management in as easy, simple and extensible a way as possible. The SPKI WG first established a list of things one might want to do with certificates (attached at the end of this document), and then summarized that list of desires into requirements. This document presents that summary of requirements. Charter of the SPKI working group Many Internet protocols and applications which use the Internet employ public key technology for security purposes and require a public key infrastructure to manage public keys. The task of the working group will be to develop Internet standards for an IETF sponsored public key certificate format, associated signature and other formats, and key acquisition protocols. The key certificate format and associated protocols are to be simple to understand, implement, and use. For purposes of the working group, the resulting formats and protocols are to be known as the Simple Public Key Infrastructure, or SPKI. The SPKI is intended to provide mechanisms to support security in a wide range of internet applications, including IPSEC protocols, encrypted electronic mail and WWW documents, payment protocols, and any other application which will require the use of public key certificates and the ability to access them. It is intended that the Simple Public Key Infrastructure will support a range of trust models. Ellison [Page 2] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 Table Of Contents Status of This Document....................................1 Abstract...................................................2 Charter of the SPKI working group..........................2 Table Of Contents..........................................3 Background.................................................4 General Requirements.......................................6 Keyholder..................................................7 Validity and CRLs..........................................7 Implementation of Certificates.............................8 List of Certificate Uses..................................10 Open Questions............................................16 References................................................17 Author's Address..........................................17 Expiration and File Name..................................17 Ellison [Page 3] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 Background The term certificate traces back to the MIT bachelor's thesis of Loren M. Kohnfelder [KOHN]. Kohnfelder, in turn, was responding to a suggestion by Diffie and Hellman in their seminal paper [DH]. Diffie and Hellman noted that with true public key cryptography, one no longer needs a secure channel over which to transmit secret keys between communicants. Instead, one can publish a modified telephone book -- one with public keys in place of telephone numbers. Diffie and Hellman went on to propose that such a directory could be on-line and maintained by a trusted source. One could then look up his or her desired communication partner in the directory, find that person's public key and open a secure channel to that person. Kohnfelder took that suggestion and noted that an on-line service has the disadvantage of being a performance bottleneck. To replace it, he proposed creation of digitally signed directory entries which he called certificates. Because these had the same content as the original proposal, they were limited to name and public key. In the time since 1978, the term certificate has frequently been assumed to mean a binding between name and key. The SPKI team directly addressed the issue of bindings and realized that such certificates are of extremely limited use for trust management. The problem is that the original telephone directory model is inappropriate for a public key infrastructure [PKI]. What a telephone directory tells you is the number for each person of a given name who desires to be listed. Imagine entering a small town where your old friend, Bill Smith, once lived. The phone book of that small town might list 5 people of that name. In order to find if your old friend is one of those and, if so, which one, you must call them all. The phone book never claimed to list your old friends. That is a function for your personal address book. In general, the relationships between people are not listed in a phone book because there are too many possible relationships. Neither does a phone book list a person's credit cards, or checking account, or habit of paying bills on time, or permission to telnet in through a firewall, or any other attribute which might be important to an on- line entity engaged in trust management. Some such information is confidential besides and therefore something the subject would not want to have published. On the other hand, the announced purpose of a PKI is to give someone the information needed to permit the extending of trust without resorting to on-line exchanges. The telephone book model does not satisfy that announced purpose. The more global the directory, the more ambiguous its names become and therefore the less well it performs. More to the point for SPKI is that for many areas of trust management a person's name is irrelevant. A user of a certificate needs to know Ellison [Page 4] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 whether a given keyholder has been granted some attribute and that attribute rarely involves a name. Names are frequently considered necessary because of the habit in the physical world of granting trust to people one knows and not to strangers. However, knowing a person's name is not the same as knowing the person. These used to be equivalent, in days when communities were small and rarely changed, when a person could live an entire lifetime and not meet anyone who had changed names, when everyone a person would meet would know the same other people and know them by the same names, etc. In that kind of world, the one in which our social habits evolved, knowing someone's name was roughly equivalent to knowing the person. In the global world of the Internet, there is no such equivalence. Ellison [Page 5] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 General Requirements The main purpose of an SPKI certificate is to authorize some action, give permission, grant a capability, etc. The first requirement for an SPKI certificate is then to bind a meaningful or useful attribute to a public key (and therefore to the keyholder of the corresponding private key). In many cases, the attribute would not involve any recognizable name. The definition of attributes or authorizations in a certificate is up to the author of code which uses the certificate. The creation of new authorizations should not require interaction with any other person or organization but rather be under the total control of the author of the code using the certificate. Because SPKI certificates might carry information which the certificate holder might not want to publish, we assume that certificates will be distributed directly by the holder to the verifier. If the holder wishes to use a global repository, similar to the global PGP key server or the DNS database, that is up to the certificate holder and not for the SPKI WG to specify. Because SPKI certificates will carry information which, taken together over all certificates, might constitute a dossier and therefore a privacy violation, each SPKI certificate should carry the minimum attribute necessary to get a job done. The SPKI certificate is then to be like a single key rather than a key ring -- a single credit card rather than a whole wallet. The certificate holder should be able to release a minimum of information in order to prove his or her permission to act. When possible, all certificates should be anonymous. Because one use of SPKI certificates is in secret balloting and similar applications, an SPKI certificate must be able to assign an attribute to a blinded signature key. For SPKI purposes, a certificate is a digitally signed testimony to whom it may concern, stating some fact or granting some permission. It is specifically not limited to a binding between name and key. However, one attribute of concern, especially for encrypted e-mail uses, is a keyholder's identity. Since humans are required to use names for things in order to think about them, people use names to think about other people. However, each person has his or her own names for friends and therefore his or her own, local namespace. For the sake of brevity, let us call names in a local namespace nicknames. An SPKI certificate must be able to bind a key to a person's Ellison [Page 6] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 nickname. The SDSI work of Rivest and Lampson has done an especially good job of defining and using local namespaces, therefore if possible SPKI should support the SDSI name construct. Keyholder Central to the SPKI effort is the notion of a keyholder. A keyholder is a person (or device) which possesses (can use) a private key. The keyholder is intended to be a single person or entity. The SPKI certificate should be designed to encourage keyholders never to share their private keys. Therefore, a certificate holder should be able to delegate permissions he acquires through the certificate. Although we humans are fond of our common names, a common name is not appropriate as a name for a keyholder. A common name does not uniquely specify most keyholders. A common name must be extended to make it unique, e.g., via a unique number. One unique number especially appropriate for identifying a keyholder is the public key. Another unique number for identifying the keyholder is a cryptographic hash of the public key. From the point of view of a verification task, the combination of common name and public key is no more unique than the public key alone, so the common name can be eliminated from this pair, leaving just the public key as the name of the keyholder. Each keyholder must be identified uniquely, with the public key itself a strong possibility for that identification. Validity and CRLs An SPKI certificate, like any other, should be able to carry a validity period: dates within which it is valid. However, the notion of a CRL has been extended by the SPKI working group. In particular, the CRL and KRL of other certificate designs speaks in the negative. The same information can be delivered in a positive statement: a periodic revalidation of a certificate or a key. Similarly, such information can be generated periodically and broadcast or made available on request and given a limited lifetime. It must be possible to specify, for each certificate, not only activity dates but also conditions which must be tested on-line or Ellison [Page 7] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 through other signed instruments. These might include lists of revoked certificates [CRLs], lists of compromised keys or periodic re-validation of certificates or keys. Any CRL, KRL or revalidation instrument must have its own lifetime. A lifetime of 0 is not possible because of communication delays and clock skews, although one can consider an instrument whose lifetime is "one use" and which is delivered only as part of a challenge/response protocol. Implementation of Certificates The authorization certificates which are envisioned for SPKI (and needed to meet the demands of the list given at the end of this document) should be generated by any user of certificates and potentially any holder of certificates. The code to generate certificates should be written by many different developers, frequently persons acting alone, operating out of garages or dorm rooms. This leads to a number of constraints on the structure and encoding of certificates. In addition, SPKI certificates should be useable in very constrained environments, such as smartcards. The code to process them and the memory to store them should both be as small as possible. An SPKI certificate should be as simple as possible. There should be a bare minimum of fields necessary to get the job done and there should be an absolute minimum (hopefully 0) of optional fields. In particular, the structure should be specific enough that the creator of a certificate is constrained by the structure definition, not by complaints (or error messages) from the reader of a certificate. An SPKI certificate should be described in as simple a method as possible, relating directly to the kind of structures a C or PASCAL programmer would normally write. No library code should be required for the packing or parsing of SPKI certificates. In particular, ASN.1 is not to be used. A certificate should be signed exactly as it is transmitted. There should be no reformatting called for in the process of checking a certificate's signature (although one might canonicalize white space during certificate input, for example, if the format is text). For efficiency, if possible, an SPKI certificate should be encoded in an LR(0) grammar. That is, neither packing nor parsing of the structure should require a scan of the data. Data should be read Ellison [Page 8] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 into the kind of structure a programmer would want to use without touching the incoming bytes more than once. For efficiency, if possible, an SPKI certificate should be packed and parsed without any recursion. Ellison [Page 9] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 List of Certificate Uses The list below is a brainstorming list, accumulated on the SPKI mailing list, of uses of such certificates. - I need a certificate to give me permission to write electronic checks. - My bank would need a certificate, proving to others that it is a bank capable of cashing electronic checks and permitted to give permission to people to write electronic checks. - My bank would issue a certificate signing the key of a master bank certifier -- perhaps NACHA -- so that I could follow a certificate chain from a key I know (my bank's) to the key of any other bank in the US and, similarly, to any other bank in the world. - I might generate a certificate (a ``reputation voucher'') for a friend to introduce him to another friend -- in which certificate I could testify to my friend's political opinion (e.g., libertarian cypherpunk) or physical characteristics or anything else of interest. - I might have a certificate giving my security clearance, signed by a governmental issuing authority. - I want a certificate for some software I have downloaded and am considering running on my computer -- to make sure it hasn't changed and that some reputable company or person stands behind it. [Douglas Barnes ] - I need certificates to bind names to public keys: - [traditional certificate] binding a key to a name, implying "all the attributes of the real person having this name are transferred to this key by this certificate". This requires unique identification of a person (which is difficult in non-digital space, as it is) and someone trustworthy binding that unique name to the key in question. In this model, a key starts out naked and acquires attributes, permissions and authority from the person bound to it. - [direct certificate] binding a name to a key, implying "I (the person who is able to use the associated private key Ellison [Page 10] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 to make this signature) declare that I go by the name of XXXXXXX." The unique identification of the key is automatic -- from the key itself or a cryptographic hash of the key. The binding is done by the key itself -- in a self-signed certificate. In this model, a key is loaded with attributes, permissions and authority directly by other certificates, not indirectly through some person's name, and this certificate declares only a name or nickname by which the key's owner likes to be addressed. - [personal binding] binding a key to a nickname. This kind of certificate is signed by me, singing someone else's key and binding it to a nickname by which I know that person. It is for my use only -- never given out -- and is a signed certificate to prevent tampering with my own private directory of keys. It says nothing about how I certified the binding to my own satisfaction between the key and my friend. - I might be doing geneology and be collecting what amounts to 3x5 cards with facts to be linked together. Some of these links would be from one content to another reference [e.g., indexing and cross-referencing]. Others might be links to the researcher who collected the fact. By rights, the fact should be signed by that researcher. Viewing only the signature on the fact and the link to the researcher, this electronic 3x5 card becomes a certificate. [Ben Laurie ] - I want to sign a contract to buy a house. What kind of certificate do I need? - I have found someone on the net and she sounds really nice. Things are leading up to cybersex. How do I make sure she's not really some 80-year-old man in a nursing home? - I have met someone on the net and would like a picture of her and her height, weight and other measurements from a trustworthy source. - Can I have a digital marriage license? - Can I have a digital divorce decree? - ..a digital Voter Registration Card? - There are a number of cards one carries in a typical wallet which could become certificates attached to a public key: Ellison [Page 11] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 - health insurance card - prescription drug card - driver's license (for permission to drive) - driver's license (for permission to buy alcohol) - supermarket discount card - supermarket check-cashing card [I know -- anachronism] - Blockbuster Video rental card - ATM card - Credit card - membership card in the ACLU, NRA, Republican party, Operation Rescue, NARAL, ACM, IEEE, ICAR.... - Red Cross blood donor card - Starbuck's Coffee buy-10-get-1-free card - DC Metro fare card - Phone calling card - Alumni Association card - REI Membership card - Car insurance card - claim check for a suitcase - claim check for a pawned radio - authorization for followup visits to a doctor, after surgery - Better Business Bureau [BBB] style reputation certificates [testimonies from satisfied customers] - BBB-style certificate that no complaints exist against a business or doctor or dentist, etc. - LDS Temple Recommend Ellison [Page 12] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 - Stock certificate - Stock option - Car title - deed to land - proof of ownership of electronic equipment with an ID number - time card certificate [activating a digital time clock] - proof of degree earned [PhD, LLD, MD, ...] - permission to write digitally signed prescriptions for drugs - permission to spend up to $X of a company's money - permission to issue nuclear launch codes - I'm a sysadmin, I want to carry a certificate, signed by SAGE, that says I'm good at the things sysadmins are good at. [marcus (m.d.) leech ] - I'm that same sysadmin, I want an ephemeral certificate that grants me root access to certain systems for the day, or the week, or... [marcus (m.d.) leech ] Certain applications *will* want some form of auditing, but the audit identity should be in the domain of the particular application... For instance an "is a system administrator of this host" certificate would probably want to include an audit identity, so you can figure out which of your multiple admins screwed something up. [Bill Sommerfeld ] - I'm an amateur radio operator. I want a signed certificate that says I'm allowed to engage in amateur radio, issued by the DOC. [I currently have a paper version of one]. This would be useful in enforcing access policies to the amateur spectrum; and in tracking abuse of that same spectrum. Heck! extend this concept to all licensed spectrum users. [marcus (m.d.) leech ] - I'm the a purchasing agent for a large corporation. I want to posses a certificate that tells our suppliers that I'm authorized to make purchases up to $15,000. I don't want the suppliers to know my name, lest their sales people bug me too Ellison [Page 13] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 much. I don't want to have to share a single "Megacorp Purchasing Department Certificate" with others doing the same job [the private key would need to be shared--yuck!]. [marcus (m.d.) leech ] - "This signed-key should be considered equivalent to the certifying-key until this certificate expires for the following purposes ..." [This is desirable when you wish to reduce the exposure of long-term keys. One way to do this is to use smartcards, but those typically have slow processors and are connected through low-bandwidth links; however, if you only use the smartcard at "login" time to certify a short-term keypair, you get high performance and low exposure of the long term key. I'll note here that this flies in the face of attempts to prevent delegation of certain rights.. Maybe we need a "delegation-allowed" bit -- but there's nothing to stop someone who wishes to delegate against the rules from also loaning out their private key..]. [Bill Sommerfeld ] - "I am the current legitimate owner of a particular chunk of internet address space." [I'd like to see ipsec eventually become usable, at least for privacy, without need for prior arrangement between sites, but I think there's a need for a "I own this address"/"I own this address range" certificate in order for ipsec to coexist with existing ip-address-based firewalls] [Bill Sommerfeld ] - "I am the current legitimate owner of a this DNS name or subtree." [Bill Sommerfeld ] - "I am the legitimate receiver of mail sent to this rfc822 email address. [this might need to be signed by a key which itself had been certified by the appropriate "DNS name owner" certificate]." [This is in case I know someone owns a particular e-mail address but I don't know their key.] [Bill Sommerfeld ] - Encryption keys for E-mail and file encryption [Tatu Ylonen ] - Authentication of people or other entities [Tatu Ylonen ] Ellison [Page 14] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 - Digital signatures (unforgeability) [Tatu Ylonen ] - Timestamping / notary services [Tatu Ylonen ] - Host authentication [Tatu Ylonen ] - Service authentication [Tatu Ylonen ] Other requirements: [Tatu Ylonen ] - Trust model must be a web (people want to choose whom they trust). People must be able to choose whome they trust or consider reliable roots (maybe with varying reliabilities). - Some applications (e.g., notary services) require highly trusted keys; generation complexity is not an issue here - Some applications (e.g., host authentication) require extremely light (or no) bureauchracy. Even communication with the central administrator may be a problem. - Especially in lower-end applications (e.g. host authentication) the people generating the keys (e.g., administrators) will change, and you will no longer want them to be able to certify. On the other hand, you will usually also not want all keys they have generated to expire. This may imply a "certification right expiration" certificate requirement, probably to be implemented together with notary services. - Keys will need to be cached locally to avoid long delays fetching frequently used keys. Cf. current name servers. The key infrastructure may in future get used almost as often as the name server. The caching and performance requirements are similar. - Reliable distribution of key revocations and other certificates (e.g., the ceasing of the right to make new certificates). May involve goals like "will have spread everywhere in 24 hours" or something like that. This interacts with caching. Ellison [Page 15] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 Open Questions Given such certificates, there remain some questions, most to do with proofs of the opposite of what a certificate is designed to do. These do not have ready answers. [Some people believe that a global namespace and traditional ID certificate might answer some of these needs, but that assumes a single, unique name assigned to each individual with no possibility of name change. A traditional ID certificate structure does not answer these needs if names are not permanent, unique and limited to one per person. For example, that unique name might be a number tattooed on a person -- or the moral equivalent. It is unlikely that any democratic government would approve the creation of such names.] - Someone digitally signs a threatening e-mail message with my private key and sends it to president@whitehouse.gov. How do I prove that I didn't compose and send the message? What kind of certificate characteristic might help me in this? - Can certificates help do a title scan for purchase of a house? - Can a certificate be issued to guarantee that I am not already married, so that I can then get a digital marriage license? - The assumption in most certificates is that the proper user will protect his private key very well, to prevent anyone else from accessing his funds. However, in some cases the certificate itself might have monitary value [permission to prescribe drugs, permission to buy alcohol, ...]. What is to prevent the holder of such a certificate from loaning out his private key? [Thanks to Bob Jueneman for this one and some of the others.] Ellison [Page 16] INTERNET-DRAFT SPKI Requirements 17 Mar 1997 References [DH] Diffie and Hellman, "New Directions in Cryptography", IEEE Transactions on Information Theory IT-22, 6 (Nov. 1976), 644-654 [KOHN] Loren Kohnfelder, "Towards a Practical Public-key Cryptosystem", Bachelor's thesis, MIT, May, 1978 Author's Address Carl M. Ellison CyberCash, Inc. 207 Grindall Street Baltimore MD 21230-4103 USA Telephone: +1 410-727-4288 +1 410-727-4293(fax) +1 703-620-4200(main office, Reston, Virginia, USA) EMail: cme@cybercash.com Expiration and File Name This draft expires 22 September 1997. Its file name is draft-ietf-spki-cert-req-00.txt Ellison [Page 17]