Network Working Group                                         C. Ellison

Request for Comments: 2693                                         Intel

Category: Experimental                                         B. Frantz

                                                    Electric Communities

                                                              B. Lampson


                                                               R. Rivest

                                     MIT Laboratory for Computer Science

                                                               B. Thomas

                                                       Southwestern Bell

                                                               T. Ylonen


                                                          September 1999



                        SPKI Certificate Theory

Note: This web page was converted automatically from a Word original. There may be problems with the formatting and the pictures. To see the intended form, follow one of the links below.


Citation: Internet RFC 2693,, Sept. 1999.

Links: Abstract, Postscript, Acrobat, Web page, Word.

Email: This paper is at


Status of this Memo


   This memo defines an Experimental Protocol for the Internet

   community.  It does not specify an Internet standard of any kind.

   Discussion and suggestions for improvement are requested.

   Distribution of this memo is unlimited.


Copyright Notice


   Copyright (C) The Internet Society (1999).  All Rights Reserved.




   The SPKI Working Group has developed a standard form for digital

   certificates whose main purpose is authorization rather than

   authentication.  These structures bind either names or explicit

   authorizations to keys or other objects.  The binding to a key can be

   directly to an explicit key, or indirectly through the hash of the

   key or a name for it.  The name and authorization structures can be

   used separately or together.  We use S-expressions as the standard

   format for these certificates and define a canonical form for those

   S-expressions.  As part of this development, a mechanism for deriving

   authorization decisions from a mixture of certificate types was

   developed and is presented in this document.


   This document gives the theory behind SPKI certificates and ACLs

   without going into technical detail about those structures or their



Table of Contents


   1. Overview of Contents.......................................3

   1.1 Glossary..................................................4

   2. Name Certification.........................................5

   2.1 First Definition of CERTIFICATE...........................6

   2.2 The X.500 Plan and X.509..................................6

   2.3 X.509, PEM and PGP........................................7

   2.4 Rethinking Global Names...................................7

   2.5 Inescapable Identifiers...................................9

   2.6 Local Names..............................................10

   2.6.1 Basic SDSI Names.......................................10

   2.6.2 Compound SDSI Names....................................10

   2.7 Sources of Global Identifiers............................11

   2.8 Fully Qualified SDSI Names...............................11

   2.9 Fully Qualified X.509 Names..............................12

   2.10 Group Names.............................................12

   3. Authorization.............................................12

   3.1 Attribute Certificates...................................13

   3.2 X.509v3 Extensions.......................................13

   3.3 SPKI Certificates........................................14

   3.4 ACL Entries..............................................15

   4. Delegation................................................15

   4.1 Depth of Delegation......................................15

   4.1.1 No control.............................................15

   4.1.2 Boolean control........................................16

   4.1.3 Integer control........................................16

   4.1.4 The choice: boolean....................................16

   4.2 May a Delegator Also Exercise the Permission?............17

   4.3 Delegation of Authorization vs. ACLs.....................17

   5. Validity Conditions.......................................18

   5.1 Anti-matter CRLs.........................................18

   5.2 Timed CRLs...............................................19

   5.3 Timed Revalidations......................................20

   5.4 Setting the Validity Interval............................20

   5.5 One-time Revalidations...................................20

   5.6 Short-lived Certificates.................................21

   5.7 Other possibilities......................................21

   5.7.1 Micali's Inexpensive On-line Results...................21

   5.7.2 Rivest's Reversal of the CRL Logic.....................21

   6. Tuple Reduction...........................................22

   6.1 5-tuple Defined..........................................23

   6.2 4-tuple Defined..........................................24

   6.3 5-tuple Reduction Rules..................................24

   6.3.1 AIntersect.............................................25

   6.3.2 VIntersect.............................................27

   6.3.3 Threshold Subjects.....................................27

   6.3.4 Certificate Path Discovery.............................28

   6.4 4-tuple Reduction........................................28

   6.4.1 4-tuple Threshold Subject Reduction....................29

   6.4.2 4-tuple Validity Intersection..........................29

   6.5 Certificate Translation..................................29

   6.5.1 X.509v1................................................29

   6.5.2 PGP....................................................30

   6.5.3 X.509v3................................................30

   6.5.4 X9.57..................................................30

   6.5.5 SDSI 1.0...............................................30

   6.5.6 SPKI...................................................31

   6.5.7 SSL....................................................31

   6.6 Certificate Result Certificates..........................32

   7. Key Management............................................33

   7.1 Through Inescapable Names................................33

   7.2 Through a Naming Authority...............................33

   7.3 Through <name,key> Certificates..........................34

   7.4 Increasing Key Lifetimes.................................34

   7.5 One Root Per Individual..................................35

   7.6 Key Revocation Service...................................36

   7.7 Threshold ACL Subjects...................................36

   8. Security Considerations...................................37



   Authors' Addresses...........................................41

   Full Copyright Statement.....................................43


1. Overview of Contents


   This document contains the following sections:


   Section 2: history of name certification, from 1976 on.


   Section 3: discussion of authorization, rather than authentication,

   as the desired purpose of a certificate.


   Section 4: discussion of delegation.


   Section 5: discussion of validity conditions: date ranges, CRLs, re-

   validations and one-time on-line validity tests.


   Section 6: definition of 5-tuples and their reduction.


   Section 7: discussion of key management.


   Section 8: security considerations.


   The References section lists all documents referred to in the text as

   well as readings which might be of interest to anyone reading on this



   The Acknowledgements section, including a list of contributors

   primarily from the start of the working group.  [The archive of

   working group mail is a more accurate source of contributor



   The Authors' Addresses section gives the addresses, telephone numbers

   and e-mail addresses of the authors.


1.1 Glossary


   We use some terms in the body of this document in ways that could be

   specific to SPKI:


   ACL: an Access Control List: a list of entries that anchors a

   certificate chain.  Sometimes called a "list of root keys", the ACL

   is the source of empowerment for certificates.  That is, a

   certificate communicates power from its issuer to its subject, but

   the ACL is the source of that power (since it theoretically has the

   owner of the resource it controls as its implicit issuer).  An ACL

   entry has potentially the same content as a certificate body, but has

   no Issuer (and is not signed).  There is most likely one ACL for each

   resource owner, if not for each controlled resource.


   CERTIFICATE: a signed instrument that empowers the Subject.  It

   contains at least an Issuer and a Subject.  It can contain validity

   conditions, authorization and delegation information.  Certificates

   come in three categories: ID (mapping <name,key>), Attribute (mapping

   <authorization,name>), and Authorization (mapping

   <authorization,key>).  An SPKI authorization or attribute certificate

   can pass along all the empowerment it has received from the Issuer or

   it can pass along only a portion of that empowerment.


   ISSUER: the signer of a certificate and the source of empowerment

   that the certificate is communicating to the Subject.


   KEYHOLDER: the person or other entity that owns and controls a given

   private key.  This entity is said to be the keyholder of the keypair

   or just the public key, but control of the private key is assumed in

   all cases.


   PRINCIPAL: a cryptographic key, capable of generating a digital

   signature.  We deal with public-key signatures in this document but

   any digital signature method should apply.


   SPEAKING: A Principal is said to "speak" by means of a digital

   signature.  The statement made is the signed object (often a

   certificate).  The Principal is said to "speak for" the Keyholder.


   SUBJECT: the thing empowered by a certificate or ACL entry.  This can

   be in the form of a key, a name (with the understanding that the name

   is mapped by certificate to some key or other object), a hash of some

   object, or a set of keys arranged in a threshold function.


   S-EXPRESSION: the data format chosen for SPKI/SDSI.  This is a LISP-

   like parenthesized expression with the limitations that empty lists

   are not allowed and the first element in any S-expression must be a

   string, called the "type" of the expression.


   THRESHOLD SUBJECT: a Subject for an ACL entry or certificate that

   specifies K of N other Subjects.  Conceptually, the power being

   transmitted to the Subject by the ACL entry or certificate is

   transmitted in (1/K) amount to each listed subordinate Subject.  K of

   those subordinate Subjects must agree (by delegating their shares

   along to the same object or key) for that power to be passed along.

   This mechanism introduces fault tolerance and is especially useful in

   an ACL entry, providing fault tolerance for "root keys".


2. Name Certification


   Certificates were originally viewed as having one function: binding

   names to keys or keys to names.  This thought can be traced back to

   the paper by Diffie and Hellman introducing public key cryptography

   in 1976.  Prior to that time, key management was risky, involved and

   costly, sometimes employing special couriers with briefcases

   handcuffed to their wrists.


   Diffie and Hellman thought they had radically solved this problem.

   "Given a system of this kind, the problem of key distribution is

   vastly simplified.  Each user generates a pair of inverse

   transformations, E and D, at his terminal.  The deciphering

   transformation, D, must be kept secret but need never be communicated

   on any channel.  The enciphering key, E, can be made public by

   placing it in a public directory along with the user's name and

   address.  Anyone can then encrypt messages and send them to the user,

   but no one else can decipher messages intended for him." [DH]


   This modified telephone book, fully public, took the place of the

   trusted courier.  This directory could be put on-line and therefore

   be available on demand, worldwide.  In considering that prospect,

   Loren Kohnfelder, in his 1978 bachelor's thesis in electrical

   engineering from MIT [KOHNFELDER], noted: "Public-key communication

   works best when the encryption functions can reliably be shared among

   the communicants (by direct contact if possible).  Yet when such a

   reliable exchange of functions is impossible the next best thing is

   to trust a third party.  Diffie and Hellman introduce a central

   authority known as the Public File."


2.1 First Definition of CERTIFICATE


   Kohnfelder then noted, "Each individual has a name in the system by

   which he is referenced in the Public File.  Once two communicants

   have gotten each other's keys from the Public File they can securely

   communicate.  The Public File digitally signs all of its

   transmissions so that enemy impersonation of the Public File is

   precluded."  In an effort to prevent performance problems, Kohnfelder

   invented a new construct: a digitally signed data record containing a

   name and a public key.  He called this new construct a CERTIFICATE.

   Because it was digitally signed, such a certificate could be held by

   non-trusted parties and passed around from person to person,

   resolving the performance problems involved in a central directory.


2.2 The X.500 Plan and X.509


   Ten years after Kohnfelder's thesis, the ISO X.509 recommendation was

   published as part of X.500.  X.500 was to be a global, distributed

   database of named entities: people, computers, printers, etc.  In

   other words, it was to be a global, on-line telephone book.  The

   organizations owning some portion of the name space would maintain

   that portion and possibly even provide the computers on which it was

   stored.  X.509 certificates were defined to bind public keys to X.500

   path names (Distinguished Names) with the intention of noting which

   keyholder had permission to modify which X.500 directory nodes.  In

   fact, the X.509 data record was originally designed to hold a

   password instead of a public key as the record-access authentication



   The original X.500 plan is unlikely ever to come to fruition.

   Collections of directory entries (such as employee lists, customer

   lists, contact lists, etc.) are considered valuable or even

   confidential by those owning the lists and are not likely to be

   released to the world in the form of an X.500 directory sub-tree.

   For an extreme example, imagine the CIA adding its directory of

   agents to a world-wide X.500 pool.


   The X.500 idea of a distinguished name (a single, globally unique

   name that everyone could use when referring to an entity) is also not

   likely to occur.  That idea requires a single, global naming

   discipline and there are too many entities already in the business of

   defining names not under a single discipline.  Legacy therefore

   militates against such an idea.

2.3 X.509, PEM and PGP


   The Privacy Enhanced Mail [PEM] effort of the Internet Engineering

   Task Force [RFC1114] adopted X.509 certificates, but with a different

   interpretation.  Where X.509 was originally intended to mean "the

   keyholder may modify this portion of the X.500 database", PEM took

   the certificate to mean "the key speaks for the named person".  What

   had been an access control instrument was now an identity instrument,

   along the lines envisioned by Diffie, Hellman and Kohnfelder.


   The insistence on X.509 certificates with a single global root

   delayed PEM's adoption past its window of viability.  RIPEM, by Mark

   Riordan of MSU, was a version of PEM without X.509 certificates.  It

   was distributed and used by a small community, but fell into disuse.

   MOSS (a MIME-enhanced version of PEM, produced by TIS (

   made certificate use optional, but received little distribution.


   At about the same time, in 1991, Phil Zimmermann's PGP was introduced

   with a different certificate model.  Instead of waiting for a single

   global root and the hierarchy of Certificate Authorities descending

   from that root, PGP allowed multiple, (hopefully) independent but not

   specially trusted individuals to sign a <name,key> association,

   attesting to its validity.  The theory was that with enough such

   signatures, that association could be trusted because not all of

   these signer would be corrupt.  This was known as the "web of trust"

   model.  It differed from X.509 in the method of assuring trust in the

   <name,key> binding, but it still intended to bind a globally unique

   name to a key.  With PEM and PGP, the intention was for a keyholder

   to be known to anyone in the world by this certified global name.


2.4 Rethinking Global Names


   The assumption that the job of a certificate was to bind a name to a

   key made sense when it was first published.  In the 1970's, people

   operated in relatively small communities.  Relationships formed face

   to face.  Once you knew who someone was, you often knew enough to

   decide how to behave with that person.  As a result, people have

   reduced this requirement to the simply stated: "know who you're

   dealing with".


   Names, in turn, are what we humans use as identifiers of persons.  We

   learn this practice as infants.  In the family environment names work

   as identifiers, even today.  What we learn as infants is especially

   difficult to re-learn later in life.  Therefore, it is natural for

   people to translate the need to know who the keyholder is into a need

   to know the keyholder's name.


   Computer applications need to make decisions about keyholders.  These

   decisions are almost never made strictly on the basis of a

   keyholder's name.  There is some other fact about the keyholder of

   interest to the application (or to the human being running the

   application).  If a name functions at all for security purposes, it

   is as an index into some database (or human memory) of that other

   information.  To serve in this role, the name must be unique, in

   order to serve as an index, and there must be some information to be



   The names we use to identify people are usually unique, within our

   local domain, but that is not true on a global scale.  It is

   extremely unlikely that the name by which we know someone, a given

   name, would function as a unique identifier on the Internet.  Given

   names continue to serve the social function of making the named

   person feel recognized when addressed by name but they are inadequate

   as the identifiers envisioned by Diffie, Hellman and Kohnfelder.


   In the 1970's and even through the early 1990's, relationships formed

   in person and one could assume having met the keyholder and therefore

   having acquired knowledge about that person.  If a name could be

   found that was an adequate identifier of that keyholder, then one

   might use that name to index into memories about the keyholder and

   then be able to make the relevant decision.


   In the late 1990's, this is no longer true.  With the explosion of

   the Internet, it is likely that one will encounter keyholders who are

   complete strangers in the physical world and will remain so.  Contact

   will be made digitally and will remain digital for the duration of

   the relationship.  Therefore, on first encounter there is no body of

   knowledge to be indexed by any identifier.


   One might consider building a global database of facts about all

   persons in the world and making that database available (perhaps for

   a fee).  The name that indexes that database might also serve as a

   globally unique ID for the person referenced.  The database entry

   under that name could contain all the information needed to allow

   someone to make a security decision.  Since there are multiple

   decision-makers, each interested in specific information, the

   database would need to contain the union of multiple sets of

   information.  However, that solution would constitute a massive

   privacy violation and would probably be rejected as politically



   A globally unique ID might even fail when dealing with people we do

   know.  Few of us know the full given names of people with whom we

   deal.  A globally unique name for a person would be larger than the

   full given name (and probably contain it, out of deference to a

   person's fondness for his or her own name).  It would therefore not

   be a name by which we know the person, barring a radical change in

   human behavior.


   A globally unique ID that contains a person's given name poses a

   special danger.  If a human being is part of the process (e.g.,

   scanning a database of global IDs in order to find the ID of a

   specific person for the purpose of issuing an attribute certificate),

   then it is likely that the human operator would pay attention to the

   familiar portion of the ID (the common name) and pay less attention

   to the rest.  Since the common name is not an adequate ID, this can

   lead to mistakes.  Where there can be mistakes, there is an avenue

   for attack.


   Where globally unique identifiers need to be used, perhaps the best

   ID is one that is uniform in appearance (such as a long number or

   random looking text string) so that it has no recognizable sub-field.

   It should also be large enough (from a sparse enough name space) that

   typographical errors would not yield another valid identifier.


2.5 Inescapable Identifiers


   Some people speak of global IDs as if they were inescapable

   identifiers, able to prevent someone from doing evil under one name,

   changing his name and starting over again.  To make that scenario

   come true, one would have to have assignment of such identifiers

   (probably by governments, at birth) and some mechanism so that it is

   always possible to get from any flesh and blood person back to his or

   her identifier.  Given that latter mechanism, any Certificate

   Authority desiring to issue a certificate to a given individual would

   presumably choose the same, inescapable name for that certificate.  A

   full set of biometrics might suffice, for example, to look up a

   person without danger of false positive in a database of globally

   assigned ID numbers and with that procedure one could implement

   inescapable IDs.


   The use of an inescapable identifier might be possible in some

   countries, but in others (such as the US) it would meet strong

   political opposition.  Some countries have government-assigned ID

   numbers for citizens but also have privacy regulations that prohibit

   the use of those numbers for routine business.  In either of these

   latter cases, the inescapable ID would not be available for use in

   routine certificates.


   There was a concern that commercial Certificate Authorities might

   have been used to bring inescapable names into existence, bypassing

   the political process and the opposition to such names in those

   countries where such opposition is strong.  As the (name,key)

   certificate business is evolving today, there are multiple competing

   CAs each creating disjoint Distinguished Name spaces.  There is also

   no real block to the creation of new CAs.  Therefore a person is able

   to drop one Distinguished Name and get another, by changing CA,

   making these names not inescapable.


2.6 Local Names


   Globally unique names may be politically undesirable and relatively

   useless, in the world of the Internet, but we use names all the time.


   The names we use are local names.  These are the names we write in

   our personal address books or use as nicknames or aliases with e-mail

   agents.  They can be IDs assigned by corporations (e.g., bank account

   numbers or employee numbers).  Those names or IDs do not need to be

   globally unique.  Rather, they need to be unique for the one entity

   that maintains that address book, e-mail alias file or list of

   accounts.  More importantly, they need to be meaningful to the person

   who uses them as indexes.


   Ron Rivest and Butler Lampson showed with SDSI 1.0 [SDSI] that one

   can not only use local names locally, one can use local names

   globally.  The clear security advantage and operational simplicity of

   SDSI names caused us in the SPKI group to adopt SDSI names as part of

   the SPKI standard.


2.6.1 Basic SDSI Names


   A basic SDSI 2.0 name is an S-expression with two elements: the word

   "name" and the chosen name.  For example,


        george:  (name fred)


   represents a basic SDSI name "fred" in the name space defined by



2.6.2 Compound SDSI Names


   If fred in turn defines a name, for example,


        fred:  (name sam)


   then george can refer to this same entity as


        george:  (name fred sam)


2.7 Sources of Global Identifiers


   Even though humans use local names, computer systems often need

   globally unique identifiers.  Even in the examples of section 2.6.2

   above, we needed to make the local names more global and did so by

   specifying the name-space owner.


   If we are using public key cryptography, we have a ready source of

   globally unique identifiers.


   When one creates a key pair, for use in public key cryptography, the

   private key is bound to its owner by good key safeguarding practice.

   If that private key gets loose from its owner, then a basic premise

   of public key cryptography has been violated and that key is no

   longer of interest.


   The private key is also globally unique.  If it were not, then the

   key generation process would be seriously flawed and we would have to

   abandon this public key system implementation.


   The private key must be kept secret, so it is not a possible

   identifier, but each public key corresponds to one private key and

   therefore to one keyholder.  The public key, viewed as a byte string,

   is therefore an identifier for the keyholder.


   If there exists a collision-free hash function, then a collision-free

   hash of the public key is also a globally unique identifier for the

   keyholder, and probably a shorter one than the public key.


2.8 Fully Qualified SDSI Names


   SDSI local names are of great value to their definer.  Each local

   name maps to one or more public keys and therefore to the

   corresponding keyholder(s).  Through SDSI's name chaining, these

   local names become useful potentially to the whole world.  [See

   section 2.6.2 for an example of SDSI name chaining.]


   To a computer system making use of these names, the name string is

   not enough.  One must identify the name space in which that byte

   string is defined.  That name space can be identified globally by a

   public key.


   It is SDSI 1.0 convention, preserved in SPKI, that if a (local) SDSI

   name occurs within a certificate, then the public key of the issuer

   is the identifier of the name space in which that name is defined.


   However, if a SDSI name is ever to occur outside of a certificate,

   the name space within which it is defined must be identified.  This

   gives rise to the Fully Qualified SDSI Name.  That name is a public

   key followed by one or more names relative to that key.  If there are

   two or more names, then the string of names is a SDSI name chain.

   For example,


        (name (hash sha1 |TLCgPLFlGTzgUbcaYLW8kGTEnUk=|) jim therese)


   is a fully qualified SDSI name, using the SHA-1 hash of a public key

   as the global identifier defining the name space and anchoring this

   name string.


2.9 Fully Qualified X.509 Names


   An X.509 Distinguished Name can and sometimes must be expressed as a

   Fully Qualified Name.  If the PEM or original X.500 vision of a

   single root for a global name space had come true, this wouldn't be

   necessary because all names would be relative to that same one root

   key.  However, there is not now and is not likely ever to be a single

   root key.  Therefore, every X.509 name should be expressed as the



        (name <root key> <leaf name>)


   if all leaf names descending from that root are unique.  If

   uniqueness is enforced only within each individual CA, then one would

   build a Fully Qualified Name chain from an X.509 certificate chain,

   yielding the form


        (name <root key> <CA(1)> <CA(2)> ... <CA(k)> <leaf name>).


2.10 Group Names


   SPKI/SDSI does not claim to enforce one key per name.  Therefore, a

   named group can be defined by issuing multiple (name,key)

   certificates with the same name -- one for each group member.


3. Authorization


   Fully qualified SDSI names represent globally unique names, but at

   every step of their construction the local name used is presumably

   meaningful to the issuer.  Therefore, with SDSI name certificates one

   can identify the keyholder by a name relevant to someone.


   However, what an application needs to do, when given a public key

   certificate or a set of them, is answer the question of whether the

   remote keyholder is permitted some access.  That application must

   make a decision.  The data needed for that decision is almost never

   the spelling of a keyholder's name.


   Instead, the application needs to know if the keyholder is authorized

   for some access.  This is the primary job of a certificate, according

   to the members of the SPKI WG, and the SPKI certificate was designed

   to meet this need as simply and directly as possible.


   We realize that the world is not going to switch to SPKI certificates

   overnight.  Therefore, we developed an authorization computation

   process that can use certificates in any format.  That process is

   described below in section 6.


   The various methods of establishing authorization are documented

   below, briefly.  (See also [UPKI])


3.1 Attribute Certificates


   An Attribute Certificate, as defined in X9.57, binds an attribute

   that could be an authorization to a Distinguished Name.  For an

   application to use this information, it must combine an attribute

   certificate with an ID certificate, in order to get the full mapping:


        authorization -> name -> key


   Presumably the two certificates involved came from different issuers,

   one an authority on the authorization and the other an authority on

   names.  However, if either of these issuers were subverted, then an

   attacker could obtain an authorization improperly.  Therefore, both

   the issuers need to be trusted with the authorization decision.


3.2 X.509v3 Extensions


   X.509v3 permits general extensions.  These extensions can be used to

   carry authorization information.  This makes the certificate an

   instrument mapping both:


        authorization -> key




        name -> key


   In this case, there is only one issuer, who must be an authority on

   both the authorization and the name.


   Some propose issuing a master X.509v3 certificate to an individual

   and letting extensions hold all the attributes or authorizations the

   individual would need.  This would require the issuer to be an

   authority on all of those authorizations.  In addition, this

   aggregation of attributes would result in shortening the lifetime of

   the certificate, since each attribute would have its own lifetime.

   Finally, aggregation of attributes amounts to the building of a

   dossier and represents a potential privacy violation.


   For all of these reasons, it is desirable that authorizations be

   limited to one per certificate.


3.3 SPKI Certificates


   A basic SPKI certificate defines a straight authorization mapping:


        authorization -> key


   If someone wants access to a keyholder's name, for logging purposes

   or even for punishment after wrong-doing, then one can map from key

   to location information (name, address, phone, ...) to get:


        authorization -> key -> name


   This mapping has an apparent security advantage over the attribute

   certificate mapping.  In the mapping above, only the


        authorization -> key


   mapping needs to be secure at the level required for the access

   control mechanism.  The


        key -> name


   mapping (and the issuer of any certificates involved) needs to be

   secure enough to satisfy lawyers or private investigators, but a

   subversion of this mapping does not permit the attacker to defeat the

   access control.  Presumably, therefore, the care with which these

   certificates (or database entries) are created is less critical than

   the care with which the authorization certificate is issued.  It is

   also possible that the mapping to name need not be on-line or

   protected as certificates, since it would be used by human

   investigators only in unusual circumstances.


3.4 ACL Entries


   SDSI 1.0 defined an ACL, granting authorization to names.  It was

   then like an attribute certificate, except that it did not need to be

   signed or issued by any key.  It was held in local memory and was

   assumed issued by the owner of the computer and therefore of the

   resource being controlled.


   In SPKI, an ACL entry is free to be implemented in any way the

   developer chooses, since it is never communicated and therefore does

   not need to be standardized.  However, a sample implementation is

   documented, as a certificate body minus the issuer field.  The ACL

   entry can have a name as a subject, as in SDSI 1.0, or it can have a

   key as a subject.  Examples of the latter include the list of SSL

   root keys in an SSL capable browser or the file .ssh/authorized_keys

   in a user's home UNIX directory.  Those ACLs are single-purpose, so

   the individual entries do not carry explicit authorizations, but SPKI

   uses explicit authorizations so that one can use common authorization

   computation code to deal with multiple applications.


4. Delegation


   One of the powers of an authorization certificate is the ability to

   delegate authorizations from one person to another without bothering

   the owner of the resource(s) involved.  One might issue a simple

   permission (e.g., to read some file) or issue the permission to

   delegate that permission further.


   Two issues arose as we considered delegation: the desire to limit

   depth of delegation and the question of separating delegators from

   those who can exercise the delegated permission.


4.1 Depth of Delegation


   There were three camps in discussing depth of delegation: no control,

   boolean control and integer control.  There remain camps in favor of

   each of these, but a decision was reached in favor of boolean



4.1.1 No control


   The argument in favor of no control is that if a keyholder is given

   permission to do something but not the permission to delegate it,

   then it is possible for that keyholder to loan out the empowered

   private key or to set up a proxy service, signing challenges or

   requests for the intended delegate.  Therefore, the attempt to

   restrict the permission to delegate is ineffective and might back-

   fire, by leading to improper security practices.


4.1.2 Boolean control


   The argument in favor of boolean control is that one might need to

   specify an inability to delegate.  For example, one could imagine the

   US Commerce Department having a key that is authorized to declare a

   cryptographic software module exportable and also to delegate that

   authorization to others (e.g., manufacturers).  It is reasonable to

   assume the Commerce Department would not issue permission to delegate

   this further.  That is, it would want to have a direct legal

   agreement with each manufacturer and issue a certificate to that

   manufacturer only to reflect that the legal agreement is in place.


4.1.3 Integer control


   The argument in favor of integer control is that one might want to

   restrict the depth of delegation in order to control the

   proliferation of a delegated permission.


4.1.4 The choice: boolean


   Of these three, the group chose boolean control.  The subject of a

   certificate or ACL entry may exercise any permission granted and, if

   delegation is TRUE, may also delegate that permission or some subset

   of it to others.


   The no control argument has logical appeal, but there remains the

   assumption that a user will value his or her private key enough not

   to loan it out or that the key will be locked in hardware where it

   can't be copied to any other user.  This doesn't prevent the user

   from setting up a signing oracle, but lack of network connectivity

   might inhibit that mechanism.


   The integer control option was the original design and has appeal,

   but was defeated by the inability to predict the proper depth of

   delegation.  One can always need to go one more level down, by

   creating a temporary signing key (e.g., for use in a laptop).

   Therefore, the initially predicted depth could be significantly off.


   As for controlling the proliferation of permissions, there is no

   control on the width of the delegation tree, so control on its depth

   is not a tight control on proliferation.


4.2 May a Delegator Also Exercise the Permission?


   We decided that a delegator is free to create a new key pair, also

   controlled by it, and delegate the rights to that key to exercise the

   delegated permission.  Therefore, there was no benefit from

   attempting to restrict the exercise of a permission by someone

   permitted to delegate it.


4.3 Delegation of Authorization vs. ACLs


   One concern with defining an authorization certificate is that the

   function can be performed by traditional <authorization,name> ACLs

   and <name,key> ID certificates defining groups.  Such a mechanism was

   described in SDSI 1.0.  A new mechanism needs to add value or it just

   complicates life for the developer.


   The argument for delegated authorization as opposed to ACLs can be

   seen in the following example.


   Imagine a firewall proxy permitting telnet and ftp access from the

   Internet into a network of US DoD machines.  Because of the

   sensitivity of that destination network, strong access control would

   be desired.  One could use public key authentication and public key

   certificates to establish who the individual keyholder was.  Both the

   private key and the keyholder's certificates could be kept on a

   Fortezza card.  That card holds X.509v1 certificates, so all that can

   be established is the name of the keyholder.  It is then the job of

   the firewall to keep an ACL, listing named keyholders and the forms

   of access they are each permitted.


   Consider the ACL itself.  Not only would it be potentially huge,

   demanding far more storage than the firewall would otherwise require,

   but it would also need its own ACL.  One could not, for example, have

   someone in the Army have the power to decide whether someone in the

   Navy got access.  In fact, the ACL would probably need not one level

   of its own ACL, but a nested set of ACLs, eventually reflecting the

   organization structure of the entire Defense Department.


   Without the ACLs, the firewall could be implemented in a device with

   no mass storage, residing in a sealed unit one could easily hold in

   one hand.  With the ACLs, it would need a large mass storage device

   that would be accessed not only while making access control decisions

   but also for updating the ACLs.


   By contrast, let the access be controlled by authorization

   certificates.  The firewall would have an ACL with one entry,

   granting a key belonging to the Secretary of Defense the right to

   delegate all access through the firewall.  The Secretary would, in

   turn, issue certificates delegating this permission to delegate to

   each of his or her subordinates.  This process would iterate, until

   some enlisted man would receive permission to penetrate that firewall

   for some specific one protocol, but not have permission to delegate

   that permission.


   The certificate structure generated would reflect the organization

   structure of the entire Defense Department, just as the nested ACLs

   would have, but the control of these certificates (via their issuance

   and revocation) is distributed and need not show up in that one

   firewall or be replicated in all firewalls.  Each individual

   delegator of permission performs a simple task, well understood.  The

   application software to allow that delegation is correspondingly



5. Validity Conditions


   A certificate, or an ACL entry, has optional validity conditions.

   The traditional ones are validity dates: not-before and not-after.

   The SPKI group resolved, in discussion, that on-line tests of various

   kinds are also validity conditions.  That is, they further refine the

   valid date range of a certificate.  Three kinds of on-line tests are

   envisioned: CRL, re-validation and one-time.


   When validity confirmation is provided by some online test, then the

   issuer of those refinements need not be the issuer of the original

   certificate.  In many cases, the business or security model for the

   two issuers is different.  However, in SPKI, the certificate issuer

   must specify the issuer of validity confirmations.


5.1 Anti-matter CRLs


   An early form of CRL [Certificate Revocation List] was modeled after

   the news print book that used to be kept at supermarket checkout

   stands.  Those books held lists of bad checking account numbers and,

   later, bad credit card numbers.  If one's payment instrument wasn't

   listed in the book, then that instrument was considered good.


   These books would be issued periodically, and delivered by some means

   not necessarily taking a constant time.  However, when a new book

   arrived, the clerk would replace the older edition with the new one

   and start using it.  This design was suited to the constraints of the

   implementation: use of physical books, delivered by physical means.

   The book held bad account numbers rather than good ones because the

   list of bad accounts was smaller.


   An early CRL design followed this model.  It had a list of revoked

   certificate identifiers.  It also had a sequence number, so that one

   could tell which of two CRLs was more recent.  A newer CRL would

   replace an older one.


   This mode of operation is like wandering anti-matter.  When the

   issuer wants to revoke a certificate, it is listed in the next CRL to

   go out.  If the revocation is urgent, then that CRL can be released

   immediately.  The CRL then follows some dissemination process

   unrelated to the needs of the consumers of the CRL.  If the CRL

   encounters a certificate it has listed, it effectively annihilates

   that certificate.  If it encounters an older CRL, it annihilates that

   CRL also, leaving a copies of itself at the verifiers it encounters.


   However, this process is non-deterministic.  The result of the

   authorization computation is at least timing dependent.  Given an

   active adversary, it can also be a security hole.  That is, an

   adversary can prevent revocation of a given certificate by preventing

   the delivery of new CRLs.  This does not require cryptographic level

   effort, merely network tampering.


   SPKI has ruled out the use of wandering anti-matter CRLs for its

   certificates.  Every authorization computation is deterministic,

   under SPKI rules.


5.2 Timed CRLs


   SPKI permits use of timed CRLs.  That is, if a certificate can be

   referenced in a CRL, then the CRL process is subject to three



   1.  The certificate must list the key (or its hash) that will sign

       the CRL and may give one or more locations where that CRL might

       be fetched.


   2.  The CRL must carry validity dates.


   3.  CRL validity date ranges must not intersect.  That is, one may

       not issue a new CRL to take effect before the expiration of the

       CRL currently deployed.


   Under these rules, no certificate that might use a CRL can be

   processed without a valid CRL and no CRL can be issued to show up as

   a surprise at the verifier.  This yields a deterministic validity

   computation, independent of clock skew, although clock inaccuracies

   in the verifier may produce a result not desired by the issuer.  The

   CRL in this case is a completion of the certificate, rather than a

   message to the world announcing a change of mind.


   Since CRLs might get very large and since they tend to grow

   monotonically, one might want to issue changes to CRLs rather than

   full ones.  That is, a CRL might be a full CRL followed by a sequence

   of delta-CRLs.  That sequence of instruments is then treated as a

   current CRL and the combined CRL must follow the conditions listed



5.3 Timed Revalidations


   CRLs are negative statements.  The positive version of a CRL is what

   we call a revalidation.  Typically a revalidation would list only one

   certificate (the one of interest), although it might list a set of

   certificates (to save digital signature effort).


   As with the CRL, SPKI demands that this process be deterministic and

   therefore that the revalidation follow the same rules listed above

   (in section 5.2).


5.4 Setting the Validity Interval


   Both timed CRLs and timed revalidations have non-0 validity

   intervals.  To set this validity interval, one must answer the

   question: "How long are you willing to let the world believe and act

   on a statement you know to be false?"


   That is, one must assume that the previous CRL or revalidation has

   just been signed and transmitted to at least one consumer, locking up

   a time slot.  The next available time slot starts after this validity

   interval ends.  That is the earliest one can revoke a certificate one

   learns to be false.


   The answer to that question comes from risk management.  It will

   probably be based on expected monetary losses, at least in commercial



5.5 One-time Revalidations


   Validity intervals of length zero are not possible.  Since

   transmission takes time, by the time a CRL was received by the

   verifier, it would be out of date and unusable.  That assumes perfect

   clock synchronization.  If clock skew is taken into consideration,

   validity intervals need to be that much larger to be meaningful.


   For those who want to set the validity interval to zero, SPKI defines

   a one-time revalidation.


   This form of revalidation has no lifetime beyond the current

   authorization computation.  One applies for this on-line, one-time

   revalidation by submitting a request containing a nonce.  That nonce

   gets returned in the signed revalidation instrument, in order to

   prevent replay attacks.  This protocol takes the place of a validity

   date range and represents a validity interval of zero, starting and

   ending at the time the authorization computation completes.


5.6 Short-lived Certificates


   A performance analysis of the various methods of achieving fine-grain

   control over the validity interval of a certificate should consider

   the possibility of just making the original certificate short-lived,

   especially if the online test result is issued by the same key that

   issued the certificate.  There are cases in which the short-lived

   certificate requires fewer signatures and less network traffic than

   the various online test options.  The use of a short-lived

   certificate always requires fewer signature verifications than the

   use of certificate plus online test result.


   If one wants to issue short-lived certificates, SPKI defines a kind

   of online test statement to tell the user of the certificate where a

   replacement short-lived certificate might be fetched.


5.7 Other possibilities


   There are other possibilities to be considered when choosing a

   validity condition model to use.


5.7.1 Micali's Inexpensive On-line Results


   Silvio Micali has patented a mechanism for using hash chains to

   revalidate or revoke a certificate inexpensively.  This mechanism

   changes the performance requirements of those models and might

   therefore change the conclusion from a performance analysis [ECR].


5.7.2 Rivest's Reversal of the CRL Logic


   Ron Rivest has written a paper [R98] suggesting that the whole

   validity condition model is flawed because it assumes that the issuer

   (or some entity to which it delegates this responsibility) decides

   the conditions under which a certificate is valid.  That traditional

   model is consistent with a military key management model, in which

   there is some central authority responsible for key release and for

   determining key validity.


   However, in the commercial space, it is the verifier and not the

   issuer who is taking a risk by accepting a certificate.  It should

   therefore be the verifier who decides what level of assurance he

   needs before accepting a credential.  That verifier needs information

   from the issuer, and the more recent that information the better, but

   the decision is the verifier's in the end.


   This line of thought deserves further consideration, but is not

   reflected in the SPKI structure definition.  It might even be that

   both the issuer and the verifier have stakes in this decision, so

   that any replacement validity logic would have to include inputs from



6. Tuple Reduction


   The processing of certificates and related objects to yield an

   authorization result is the province of the developer of the

   application or system.  The processing plan presented here is an

   example that may be followed, but its primary purpose is to clarify

   the semantics of an SPKI certificate and the way it and various other

   kinds of certificate might be used to yield an authorization result.


   There are three kinds of entity that might be input to the

   computation that yields an authorization result:


    1.  <name,key> (as a certificate)


    2.  <authorization,name> (as an attribute certificate or ACL entry)


    3.  <authorization,key> (as an authorization certificate or ACL



   These entities are processed in three stages.


    1.  Individual certificates are verified by checking their

        signatures and possibly performing other work.  They are then

        mapped to intermediate forms, called "tuples" here.


        The other work for SPKI or SDSI certificates might include

        processing of on-line test results (CRL, re-validation or one-

        time validation).


        The other work for PGP certificates may include a web-of-trust



        The other work for X.509 certificates depends on the written

        documentation for that particular use of X.509 (typically tied

        to the root key from which the certificate descended) and could

        involve checking information in the parent certificate as well

        as additional information in extensions of the certificate in

        question.  That is, some use X.509 certificates just to define

        names.  Others use X.509 to communicate an authorization

        implicitly (e.g., SSL server certificates).  Some might define

        extensions of X.509 to carry explicit authorizations.  All of

        these interpretations are specified in written documentation

        associated with the certificate chain and therefore with the

        root from which the chain descends.


        If on-line tests are involved in the certificate processing,

        then the validity dates of those on-line test results are

        intersected by VIntersect() [defined in 6.3.2, below] with the

        validity dates of the certificate to yield the dates in the

        certificate's tuple(s).


    2.  Uses of names are replaced with simple definitions (keys or

        hashes), based on the name definitions available from reducing

        name 4-tuples.


    3.  Authorization 5-tuples are then reduced to a final authorization



6.1 5-tuple Defined


   The 5-tuple is an intermediate form, assumed to be held in trusted

   memory so that it doesn't need a digital signature for integrity.  It

   is produced from certificates or other credentials via trusted

   software.  Its contents are the same as the contents of an SPKI

   certificate body, but it might be derived from another form of

   certificate or from an ACL entry.


   The elements of a 5-tuple are:


    1.  Issuer: a public key (or its hash), or the reserved word "Self".

        This identifies the entity speaking this intermediate result.


    2.  Subject: a public key (or its hash), a name used to identify a

        public key, the hash of an object or a threshold function of

        subordinate subjects.  This identifies the entity being spoken

        about in this intermediate result.


    3.  Delegation: a boolean.  If TRUE, then the Subject is permitted

        by the Issuer to further propagate the authorization in this

        intermediate result.


    4.  Authorization: an S-expression.  [Rules for combination of

        Authorizations are given below.]


    5.  Validity dates: a not-before date and a not-after date, where

        "date" means date and time.  If the not-before date is missing

        from the source credential then minus infinity is assumed.  If

        the not-after date is missing then plus infinity is assumed.


6.2 4-tuple Defined


   A <name,key> certificate (such as X.509v1 or SDSI 1.0) carries no

   authorization field but does carry a name.  Since it is qualitatively

   different from an authorization certificate, a separate intermediate

   form is defined for it.


   The elements of a Name 4-tuple are:


    1.  Issuer: a public key (or its hash).  This identifies the entity

        defining this name in its private name space.


    2.  Name: a byte string


    3.  Subject: a public key (or its hash), a name, or a threshold

        function of subordinate subjects.  This defines the name.


    4.  Validity dates: a not-before date and a not-after date, where

        "date" means date and time.  If the not-before date is missing

        from the source credential then minus infinity is assumed.  If

        the not-after date is missing then plus infinity is assumed.


6.3 5-tuple Reduction Rules


   The two 5-tuples:


      <I1,S1,D1,A1,V1> + <I2,S2,D2,A2,V2>








       the two intersections succeed,


       S1 = I2




       D1 = TRUE


   If S1 is a threshold subject, there is a slight modification to this

   rule, as described below in section 6.3.3.


6.3.1 AIntersect


   An authorization is a list of strings or sub-lists, of meaning to and

   probably defined by the application that will use this authorization

   for access control.  Two authorizations intersect by matching,

   element for element.  If one list is longer than the other but match

   at all elements where both lists have elements, then the longer list

   is the result of the intersection.  This means that additional

   elements of a list must restrict the permission granted.


   Although actual authorization string definitions are application

   dependent, AIntersect provides rules for automatic intersection of

   these strings so that application developers can know the semantics

   of the strings they use.  Special semantics would require special

   reduction software.


   For example, there might be an ftpd that allows public key access

   control, using authorization certificates.  Under that service,


       (ftp (host


   might imply that the keyholder would be allowed ftp access to all

   directories on, with all kinds of access (read, write,

   delete, ...).  This is more general (allows more access) than


       (ftp (host (dir /pub/cme))


   which would allow all kinds of access but only in the directory

   specified.  The intersection of the two would be the second.


   Since the AIntersect rules imply position dependency, one could also

   define the previous authorization string as:


       (ftp /pub/cme)


   to keep the form compact.


   To allow for wild cards, there are a small number of special S-

   expressions defined, using "*" as the expression name.



             stands for the set of all S-expressions and byte-strings.

             In other words, it will match anything.  When intersected

             with anything, the result is that other thing.  [The

             AIntersect rule about lists of different length treats a

             list as if it had enough (*) entries implicitly appended to

             it to match the length of another list with which it was

             being intersected.]


   (* set <tag-expr>*)

             stands for the set of elements listed in the *-form.


   (* prefix <byte-string>)

             stands for the set of all byte strings that start with the

             one given in the *-form.


   (* range <ordering> <lower-limit>? <upper-limit>?)

             stands for the set of all byte strings lexically (or

             numerically) between the two limits.  The ordering

             parameter (alpha, numeric, time, binary, date) specifies

             the kind of strings allowed.


   AIntersect() is normal set intersection, when *-forms are defined as

   they are above and a normal list is taken to mean all lists that

   start with those elements.  The following examples should give a more

   concrete explanation for those who prefer an explanation without

   reference to set operations.


   AIntersect( (tag (ftp cme (* set read write))),

               (tag (*)) )


   evaluates to (tag (ftp cme (* set read write)))


   AIntersect( (tag (* set read write (foo bla) delete)),

               (tag (* set write read) ) )


   evaluates to (tag (* set read write))


   AIntersect( (tag (* set read write (foo bla) delete)),

               (tag read ) )


   evaluates to (tag read)


   AIntersect( (tag (* prefix,

               (tag (* prefix )


   evaluates to (tag (* prefix


   AIntersect( (tag (* range numeric ge #30# le #39# )), (tag #26#) )


   fails to intersect.


6.3.2 VIntersect


   Date range intersection is straight-forward.


       V = VIntersect( X, Y )


   is defined as


       Vmin = max( Xmin, Ymin )


       Vmax = min( Xmax, Ymax )


   and if Vmin > Vmax, then the intersection failed.


   These rules assume that daytimes are expressed in a monotonic form,

   as they are in SPKI.


   The full SPKI VIntersect() also deals with online tests.  In the most

   straight-forward implementation, each online test to which a

   certificate is subject is evaluated.  Each such test carries with it

   a validity interval, in terms of dates.  That validity interval is

   intersected with any present in the certificate, to yield a new,

   current validity interval.


   It is possible for an implementation of VIntersect() to gather up

   online tests that are present in each certificate and include the

   union of all those tests in the accumulating tuples.  In this case,

   the evaluation of those online tests is deferred until the end of the

   process.  This might be appropriate if the tuple reduction is being

   performed not for answering an immediate authorization question but

   rather for generation of a summary certificate (Certificate Result

   Certificate) that one might hope would be useful for a long time.


6.3.3 Threshold Subjects


   A threshold subject is specified by two numbers, K and N [0<K<=N],

   and N subordinate subjects.  A threshold subject is reduced to a

   single subject by selecting K of the N subjects and reducing each of

   those K to the same subject, through a sequence of certificates.  The

   (N-K) unselected subordinate subjects are set to (null).


   The intermediate form for a threshold subject is a copy of the tuple

   in which the threshold subject appears, but with only one of the

   subordinate subjects.  Those subordinate tuples are reduced

   individually until the list of subordinate tuples has (N-K) (null)

   entries and K entries with the same subject.  At that point, those K

   tuples are validity-, authorization- and delegation- intersected to

   yield the single tuple that will replace the list of tuples.




Ellison, et al.               Experimental                     [Page 27]


RFC 2693                SPKI Certificate Theory           September 1999



6.3.4 Certificate Path Discovery


   All reduction operations are in the order provided by the prover.

   That simplifies the job of the verifier, but leaves the job of

   finding the correct list of reductions to the prover.


   The general algorithm for finding the right certificate paths from a

   large set of unordered certificates has been solved[ELIEN], but might

   be used only rarely.  Each keyholder who is granted some authority

   should receive a sequence of certificates delegating that authority.

   That keyholder may then want to delegate part of this authority on to

   some other keyholder.  To do that, a single additional certificate is

   generated and appended to the sequence already available, yielding a

   sequence that can be used by the delegatee to prove access



6.4 4-tuple Reduction


   There will be name 4-tuples in two different classes, those that

   define the name as a key and those that define the name as another



    1.  [(name K1 N) -> K2]


    2.  [(name K1 N) -> (name K2 N1 N2 ... Nk)]


   As with the 5-tuples discussed in the previous section, name

   definition 4-tuples should be delivered in the order needed by the

   prover.  In that case, the rule for name reduction is to replace the

   name just defined by its definition.  For example,


        (name K1 N N1 N2 N3) + [(name K1 N) -> K2]


             -> (name K2 N1 N2 N3)


   or, in case 2 above,


        (name K1 N Na Nb Nc) + [(name K1 N) -> (name K2 N1 N2 ... Nk)]


             -> (name K2 N1 N2 ... Nk Na Nb Nc)


   With the second form of name definition, one might have names that

   temporarily grow.  If the prover is providing certificates in order,

   then the verifier need only do as it is told.


   If the verifier is operating from an unordered pool of tuples, then a

   safe rule for name reduction is to apply only those 4-tuples that

   define a name as a key.  Such applications should bring 4-tuples that

   started out in class (2) into class (1), and eventually reduce all

   names to keys.  Any naming loops are avoided by this process.


6.4.1 4-tuple Threshold Subject Reduction


   Some of a threshold subject's subordinate subjects might be names.

   Those names must be reduced by application of 4-tuples.  The name

   reduction process proceeds independently on each name in the

   subordinate subject as indicated in 6.3.3 above.


   One can reduce individual named subjects in a threshold subject and

   leave the subject in threshold form, if one desires.  There is no

   delegation- or authorization-intersection involved, only a validity-

   intersection during name reduction.  This might be used by a service

   that produces Certificate Result Certificates [see 6.7].


6.4.2 4-tuple Validity Intersection


   Whenever a 4-tuple is used to reduce the subject (or part of the

   subject) of another tuple, its validity interval is intersected with

   that of the tuple holding the subject being reduced and the

   intersection is used in the resulting tuple.  Since a 4-tuple

   contains no delegation or authorization fields, the delegation

   permission and authorization of the tuple being acted upon does not



6.5 Certificate Translation


   Any certificate currently defined, as well as ACL entries and

   possibly other instruments, can be translated to 5-tuples (or name

   tuples) and therefore take part in an authorization computation.  The

   specific rules for those are given below.


6.5.1 X.509v1


   The original X.509 certificate is a <name,key> certificate.  It

   translates directly to a name tuple.  The form


        [Kroot, (name <leaf-name>), K1, validity]


   is used if the rules for that particular X.509 hierarchy is that all

   leaf names are unique, under that root.  If uniqueness of names

   applies only to individual CAs in the X.509 hierarchy, then one must



        [Kroot, (name CA1 CA2 ... CAk <leaf-name>), K1, validity]


   after verifying the certificate chain by the rules appropriate to

   that particular chain.


6.5.2 PGP


   A PGP certificate is a <name,key> certificate.  It is verified by

   web-of-trust rules (as specified in the PGP documentation).  Once

   verified, it yields name tuples of the form


        [Ki, name, K1, validity]


   where Ki is a key that signed that PGP (UserID,key) pair.  There

   would be one tuple produced for each signature on the key, K1.


6.5.3 X.509v3


   An X.509v3 certificate may be used to declare a name.  It might also

   declare explicit authorizations, by way of extensions.  It might also

   declare an implicit authorization of the form (tag (*)).  The actual

   set of tuples it yields depends on the documentation associated with

   that line of certificates.  That documentation could conceptually be

   considered associated with the root key of the certificate chain.  In

   addition, some X.509v3 certificates (such as those used for SET),

   have defined extra validity tests for certificate chains depending on

   custom extensions.  As a result, it is likely that X.509v3 chains

   will have to be validated independently, by chain validation code

   specific to each root key.  After that validation, that root-specific

   code can then generate the appropriate multiple tuples from the one



6.5.4 X9.57


   An X9.57 attribute certificate should yield one or more 5-tuples,

   with names as Subject.  The code translating the attribute

   certificate will have to build a fully-qualified name to represent

   the Distinguished Name in the Subject.  For any attribute

   certificates that refer to an ID certificate explicitly, the Subject

   of the 5-tuple can be the key in that ID certificate, bypassing the

   construction of name 4-tuples.


6.5.5 SDSI 1.0


   A SDSI 1.0 certificate maps directly to one 4-tuple.


6.5.6 SPKI


   An SPKI certificate maps directly to one 4- or 5- tuple, depending

   respectively on whether it is a name certificate or carries an



6.5.7 SSL


   An SSL certificate carries a number of authorizations, some

   implicitly.  The authorization:


        (tag (ssl))


   is implicit.  In addition, the server certificate carries a DNS name

   parameter to be matched against the DNS name of the web page to which

   the connection is being made.  That might be encoded as:


        (tag (dns <domain-name>))


   Meanwhile, there is the "global cert" permission -- the permission

   for a US-supplied browser to connect using full strength symmetric

   cryptography even though the server is outside the USA.  This might

   be encoded as:


        (tag (us-crypto))


   There are other key usage attributes that would also be encoded as

   tag fields, but a full discussion of those fields is left to the

   examples document.


   An ACL entry for an SSL root key would have the tag:


        (tag (* set (ssl) (dns (*))))


   which by the rules of intersection is equivalent to:


        (tag (* set (ssl) (dns)))


   unless that root key also had the permission from the US Commerce

   Department to grant us-crypto permission, in which case the root key

   would have:


        (tag (* set (ssl) (dns) (us-crypto)))


   A CA certificate, used for SSL, would then need only to communicate

   down its certificate chain those permissions allocated in the ACL.

   Its tag might then translate to:


        (tag (*))


   A leaf server certificate for the Datafellows server might, for

   example, have a tag field of the form:


        (tag (* set (ssl) (dns


   showing that it was empowered to do SSL and to operate from the given

   domain name, but not to use US crypto with a US browser.


   The use of (* set) for the two attributes in this example could have

   been abbreviated as:


        (tag (ssl


   while CA certificates might carry:


        (tag (ssl (*))) or just (tag (*))


   but separating them, via (* set), allows for a future enhancement of

   SSL in which the (ssl) permission is derived from one set of root

   keys (those of current CAs) while the (dns) permission is derived

   from another set of root keys (those empowered to speak in DNSSEC)

   while the (us-crypto) permission might be granted only to a root key

   belonging to the US Department of Commerce.  The three separate tests

   in the verifying code (e.g., the browser) would then involve separate

   5-tuple reductions from separate root key ACL entries.


   The fact that these three kinds of permission are treated as if ANDed

   is derived from the logic of the code that interprets the permissions

   and is not expressed in the certificate.  That decision is embodied

   in the authorization code executed by the verifying application.


6.6 Certificate Result Certificates


   Typically, one will reduce a chain of certificates to answer an

   authorization question in one of two forms:


    1.  Is this Subject, S, allowed to do A, under this ACL and with

        this set of certificates?


    2.  What is Subject S allowed to do, under this ACL and with this

        set of certificates?


   The answer to the second computation can be put into a new

   certificate issued by the entity doing the computation.  That one

   certificate corresponds to the semantics of the underlying

   certificates and online test results.  We call it a Certificate

   Result Certificate.


7. Key Management


   Cryptographic keys have limited lifetimes.  Keys can be stolen.  Keys

   might also be discovered through cryptanalysis.  If the theft is

   noticed, then the key can be replaced as one would replace a credit

   card.  More likely, the theft will not be noticed.  To cover this

   case, keys are replaced routinely.


   The replacement of a key needs to be announced to those who would use

   the new key.  It also needs to be accomplished smoothly, with a

   minimum of hassle.


   Rather than define a mechanism for declaring a key to be bad or

   replaced, SPKI defines a mechanism for giving certificates limited

   lifetimes so that they can be replaced.  That is, under SPKI one does

   not declare a key to be bad but rather stops empowering it and

   instead empowers some other key.  This limitation of a certificate's

   lifetime might be by limited lifetime at time of issuance or might be

   via the lifetime acquired through an on-line test (CRL, revalidation

   or one-time).  Therefore, all key lifetime control becomes

   certificate lifetime control.


7.1 Through Inescapable Names


   If keyholders had inescapable names [see section 2.5, above], then

   one could refer to them by those names and define a certificate to

   map from an inescapable name to the person's current key.  That

   certificate could be issued by any CA, since all CAs would use the

   inescapable name for the keyholder.  The attribute certificates and

   ACLs that refer to the keyholder would all refer to this one

   inescapable name.


   However, there are no inescapable names for keyholders.  [See section

   2.5, above.]


7.2 Through a Naming Authority


   One could conceivably have a governmental body or other entity that

   would issue names voluntarily to a keyholder, strictly for the

   purpose of key management.  One would then receive all authorizations

   through that name.  There would have to be only one such authority,

   however.  Otherwise, names would have to be composed of parts: an

   authority name and the individual's name.  The authority name would,

   in turn, have to be granted by some single global authority.


   That authority then becomes able to create keys of its own and

   certificates to empower them as any individual, and through those

   false certificates acquire access rights of any individual in the

   world.  Such power is not likely to be tolerated.  Therefore, such a

   central authority is not likely to come to pass.


7.3 Through <name,key> Certificates


   Instead of inescapable names or single-root naming authorities, we

   have names assigned by some entity that issues a <name,key>

   certificate.  As noted in sections 2.8 and 2.9, above, such names

   have no meaning by themselves.  They must be fully qualified to have



   Therefore, in the construct:


        (name (hash sha1 |TLCgPLFlGTzgUbcaYLW8kGTEnUk=|) jim)


   the name is not




   but rather


        "(name (hash sha1 |TLCgPLFlGTzgUbcaYLW8kGTEnUk=|) jim)"


   This name includes a public key (through its hash, in the example

   above).  That key has a lifetime like any other key, so this name has

   not achieved the kind of permanence (free from key lifetimes) that an

   inescapable name has.  However, it appears to be our only



   This name could easily be issued by the named keyholder, for the

   purpose of key management only.  In that case, there is no concern

   about access control being subverted by some third-party naming



7.4 Increasing Key Lifetimes


   By the logic above, any name will hang off some public key.  The job

   is then to increase the lifetime of that public key.  Once a key

   lifetime exceeds the expected lifetime of any authorization granted

   through it, then a succession of new, long-lifetime keys can cover a

   keyholder forever.


   For a key to have a long lifetime, it needs to be strong against

   cryptanalytic attack and against theft.  It should be used only on a

   trusted machine, running trusted software.  It should not be used on

   an on-line machine.  It should be used very rarely, so that the

   attacker has few opportunities to find the key in the clear where it

   can be stolen.


   Different entities will approach this set of requirements in

   different ways.  A private individual, making his own naming root key

   for this purpose, has the advantage of being too small to invite a

   well funded attack as compared to the attacks a commercial CA might



7.5 One Root Per Individual


   In the limit, one can have one highly protected naming root key for

   each individual.  One might have more than one such key per

   individual, in order to frustrate attempts to build dossiers, but let

   us assume only one key for the immediate discussion.


   If there is only one name descending from such a key, then one can

   dispense with the name.  Authorizations can be assigned to the key

   itself, in raw SPKI style, rather than to some name defined under

   that key.  There is no loss of lifetime -- only a change in the

   subject of the certificate the authorizing key uses to delegate



   However, there is one significant difference, under the SPKI

   structure.  If one delegates some authorization to


        (name (hash sha1 |TLCgPLFlGTzgUbcaYLW8kGTEnUk=|) carl)


   and a different authorization to


        (hash sha1 |TLCgPLFlGTzgUbcaYLW8kGTEnUk=|)


   directly, both without granting the permission to delegate, that key

   can delegate at will through <name,key> certificates in the former

   case and not delegate at all in the latter case.


   In the case of key management, we desire the ability to delegate from

   a long lived, rarely used key to a shorter lived, often used key --

   so in this case, the former mechanism (through a SDSI name) gives

   more freedom.


7.6 Key Revocation Service


   In either of the models above, key |TLCgPLFlGTzgUbcaYLW8kGTEnUk=|

   will issue a certificate.  In the first model, it will be a

   <name,key> certificate.  In the second, it will be an authorization

   certificate delegating all rights through to the more temporary key.


   Either of those certificates might want an on-line validity test.

   Whether this test is in the form of a CRL, a re-validation or a one-

   time test, it will be supplied by some entity that is on-line.


   As the world moves to having all machines on-line all the time, this

   might be the user's machine.  However, until then -- and maybe even

   after then -- the user might want to hire some service to perform

   this function.  That service could run a 24x7 manned desk, to receive

   phone calls reporting loss of a key.  That authority would not have

   the power to generate a new key for the user, only to revoke a

   current one.


   If, in the worst case, a user loses his master key, then the same

   process that occurs today with lost wallets would apply.  All issuers

   of authorizations through that master key would need to issue new

   authorizations through the new master key and, if the old master key

   had been stolen, cancel all old authorizations through that key.


7.7 Threshold ACL Subjects


   One can take extraordinary measures to protect root keys and thus

   increase the lifetimes of those keys.  The study of computer fault-

   tolerance teaches us that truly long lifetimes can be achieved only

   by redundancy and replacement.  Both can be achieved by the use of

   threshold subjects [section 6.3.3], especially in ACL entries.


   If we use a threshold subject in place of a single key subject, in an

   ACL (or a certificate), then we achieve redundancy immediately.  This

   can be redundancy not only of keys but also of algorithms.  That is,

   the keys in a threshold subject do not need to have the same



   Truly long lifetimes come from replacement, not just redundancy.  As

   soon as a component fails (or a key is assumed compromised), it must

   be replaced.


   An ACL needs to be access-controlled itself.  Assume that the ACL

   includes an entry with authorization


       (tag (acl-edit))


   Assume also that what might have been a single root authorization

   key, K1, is actually a threshold subject


       (k-of-n #03# #07# K1 K2 K3 K4 K5 K6 K7)


   used in any ACL entry granting a normal authorization.


   That same ACL could have the subject of an (acl-edit) entry be


       (k-of-n #05# #07# K1 K2 K3 K4 K5 K6 K7)


   This use of threshold subject would allow the set of root keys to

   elect new members to that set and retire old members.  In this

   manner, replacement is achieved alongside redundancy and the proper

   choice of K and N should allow threshold subject key lifetimes

   approaching infinity.


8. Security Considerations


   There are three classes of information that can be bound together by

   public key certificates: key, name and authorization.  There are

   therefore three general kinds of certificate, depending on what pair

   of items the certificate ties together.  If one considers the

   direction of mapping between items, there are six classes: name->key,

   key->name, authorization->name, name->authorization, authorization-

   >key, key->authorization.


   The SPKI working group concluded that the most important use for

   certificates was access control.  Given the various kinds of mapping

   possible, there are at least two ways to implement access control.

   One can use a straight authorization certificate:




   or one can use an attribute certificate and an ID certificate:


       (authorization->name) + (name->key)


   There are at least two ways in which the former is more secure than

   the latter.


    1.  Each certificate has an issuer.  If that issuer is subverted,

        then the attacker can gain access.  In the former case, there is

        only one issuer to trust.  In the latter case, there are two.


    2.  In the second case, linkage between the certificates is by name.

        If the name space of the issuer of the ID certificate is

        different from the name space of the issuer of the attribute

        certificate, then one of the two issuers must use a foreign name

        space.  The process of choosing the appropriate name from a

        foreign name space is more complex than string matching and

        might even involve a human guess.  It is subject to mistakes.

        Such a mistake can be made by accident or be guided by an



   This is not to say that one must never use the second construct.  If

   the two certificates come from the same issuer, and therefore with

   the same name space, then both of the security differentiators above

   are canceled.




   [Ab97]       Abadi, Martin, "On SDSI's Linked Local Name Spaces",

                Proceedings of the 10th IEEE Computer Security

                Foundations Workshop (June 1997).


   [BFL]        Matt Blaze, Joan Feigenbaum and Jack Lacy, "Distributed

                Trust Management", Proceedings 1996 IEEE Symposium on

                Security and Privacy.


   [CHAUM]      D. Chaum, "Blind Signatures for Untraceable Payments",

                Advances in Cryptology -- CRYPTO '82, 1983.


   [DH]         Whitfield Diffie and Martin Hellman, "New Directions in

                Cryptography", IEEE Transactions on Information Theory,

                November 1976, pp. 644-654.


   [DvH]        J. B. Dennis and E. C. Van Horn, "Programming Semantics

                for Multiprogrammed Computations", Communications of the

                ACM 9(3), March 1966.


   [ECR]        Silvio Micali, "Efficient Certificate Revocation",

                manuscript, MIT LCS.


   [ELIEN]      Jean-Emile Elien, "Certificate Discovery Using SPKI/SDSI

                2.0 Certificates", Masters Thesis, MIT LCS, May 1998,


                [also .pdf and


   [HARDY]      Hardy, Norman, "THE KeyKOS Architecture", Operating

                Systems Review, v.19 n.4, October 1985. pp 8-25.


   [IDENT]      Carl Ellison, "Establishing Identity Without

                Certification Authorities", USENIX Security Symposium,

                July 1996.


   [IWG]        McConnell and Appel, "Enabling Privacy, Commerce,

                Security and Public Safety in the Global Information

                Infrastructure", report of the Interagency Working Group

                on Cryptography Policy, May 12, 1996; (quote from

                paragraph 5 of the Introduction).


   [KEYKOS]     Bomberger, Alan, et al., "The KeyKOS(r) Nanokernel

                Architecture", Proceedings of the USENIX Workshop on

                Micro-Kernels and Other Kernel Architectures, USENIX

                Association, April 1992. pp 95-112 (In addition, there

                are KeyKOS papers on the net available through



   [KOHNFELDER] Kohnfelder, Loren M., "Towards a Practical Public-key

                Cryptosystem", MIT S.B. Thesis, May. 1978.


   [LAMPSON]    B. Lampson, M. Abadi, M. Burrows, and E. Wobber,

                "Authentication in distributed systems: Theory and

                practice", ACM Trans. Computer Systems 10, 4 (Nov.

                1992), pp 265-310.


   [LANDAU]     Landau, Charles, "Security in a Secure Capability-Based

                System", Operating Systems Review, Oct 1989 pp 2-4.


   [LEVY]       Henry M. Levy, "Capability-Based Computer Systems",

                Digital Press, 12 Crosby Dr., Bedford MA 01730, 1984.


   [LINDEN]     T. A. Linden, "Operating System Structures to Support

                Security and Reliable Software", Computing Surveys 8(4),

                December 1976.


   [PKCS1]      PKCS #1: RSA Encryption Standard, RSA Data Security,

                Inc., 3 June 1991, Version 1.4.


   [PKLOGIN]    David Kemp, "The Public Key Login Protocol", Work in



   [R98]        R. Rivest, "Can We Eliminate Revocation Lists?", to

                appear in the Proceedings of Financial Cryptography

                1998, <>.


   [RFC1114]    Kent, S. and  J. Linn, "Privacy Enhancement for Internet

                Electronic Mail: Part II -- Certificate-Based Key

                Management", RFC 1114, August 1989.


   [RFC1321]    Rivest, R., "The MD5 Message-Digest Algorithm", RFC

                1321, April 1992.


   [RFC2045]    Freed, N. and N. Borenstein, "Multipurpose Internet Mail

                Extensions (MIME) Part One: Format of Internet Message

                Bodies", RFC 2045, December 1996.


   [RFC2046]    Freed, N. and N. Borenstein, "Multipurpose Internet Mail

                Extensions (MIME) Part Two: Media Types", RFC 2046,

                December 1996.


   [RFC2047]    K. Moore, "MIME (Multipurpose Internet Mail Extensions)

                Part Three: Message Header Extensions for Non-ASCII

                Text", RFC 2047, December 1996.


   [RFC2065]    Eastlake, D. and C. Kaufman, "Proposed Standard for DNS

                Security", RFC 2065, January 1997.


   [RFC2104]    Krawczyk, H., Bellare, M. and R. Canetti, "HMAC:

                Keyed-Hashing for Message Authentication", RFC 2104,

                February 1997.


   [SDSI]       Ron Rivest and Butler Lampson, "SDSI - A Simple

                Distributed Security Infrastructure [SDSI]",



   [SET]        Secure Electronic Transactions -- a protocol designed by

                VISA, MasterCard and others, including a certificate

                structure covering all participants.  See



   [SEXP]       Ron Rivest, code and description of S-expressions,



   [SRC-070]    Abadi, Burrows, Lampson and Plotkin, "A Calculus for

                Access Control in Distributed Systems", DEC SRC-070,

                revised August 28, 1991.


   [UPKI]       C. Ellison, "The nature of a useable PKI", Computer

                Networks 31 (1999) pp. 823-830.


   [WEBSTER]    "Webster's Ninth New Collegiate Dictionary", Merriam-

                Webster, Inc., 1991.




   Several independent contributions, published elsewhere on the net or

   in print, worked in synergy with our effort.  Especially important to

   our work were: [SDSI], [BFL] and [RFC2065].  The inspiration we

   received from the notion of CAPABILITY in its various forms (SDS-940,

   Kerberos, DEC DSSA, [SRC-070], KeyKOS [HARDY]) can not be over-rated.


   Significant contributions to this effort by the members of the SPKI

   mailing list and especially the following persons (listed in

   alphabetic order) are gratefully acknowledged: Steve Bellovin, Mark

   Feldman, John Gilmore, Phill Hallam-Baker, Bob Jueneman, David Kemp,

   Angelos D. Keromytis, Paul Lambert, Jon Lasser, Jeff Parrett, Bill

   Sommerfeld, Simon Spero.


Authors' Addresses


   Carl M. Ellison

   Intel Corporation

   2111 NE 25th Ave  M/S JF3-212

   Hillsboro OR 97124-5961 USA


   Phone: +1-503-264-2900

   Fax:   +1-503-264-6225





   Bill Frantz

   Electric Communities

   10101 De Anza Blvd.

   Cupertino CA 95014


   Phone: +1 408-342-9576




   Butler Lampson


   180 Lake View Ave

   Cambridge MA 02138


   Phone: +1 617-547-9580 (voice + FAX)



   Ron Rivest

   Room 324, MIT Laboratory for Computer Science

   545 Technology Square

   Cambridge MA 02139


   Phone: +1-617-253-5880

   Fax:   +1-617-258-9738





   Brian Thomas

   Southwestern Bell

   One Bell Center, Room 34G3

   St. Louis MO 63101 USA


   Phone: +1 314-235-3141

   Fax:   +1 314-235-0162




   Tatu Ylonen

   SSH Communications Security Ltd.

   Tekniikantie 12

   FIN-02150 ESPOO






Full Copyright Statement


   Copyright (C) The Internet Society (1999).  All Rights Reserved.


   This document and translations of it may be copied and furnished to

   others, and derivative works that comment on or otherwise explain it

   or assist in its implementation may be prepared, copied, published

   and distributed, in whole or in part, without restriction of any

   kind, provided that the above copyright notice and this paragraph are

   included on all such copies and derivative works.  However, this

   document itself may not be modified in any way, such as by removing

   the copyright notice or references to the Internet Society or other

   Internet organizations, except as needed for the purpose of

   developing Internet standards in which case the procedures for

   copyrights defined in the Internet Standards process must be

   followed, or as required to translate it into languages other than



   The limited permissions granted above are perpetual and will not be

   revoked by the Internet Society or its successors or assigns.


   This document and the information contained herein is provided on an









   Funding for the RFC Editor function is currently provided by the

   Internet Society.