Public-key cryptography and related standards and techniques underlie the security features of many products such as signed and encrypted email, single sign-on, and Secure Sockets Layer (SSL) communications. This document introduces the basic concepts of public-key cryptography. For an overview of SSL, see "Introduction to SSL." For an overview of encryption and decryption, see "Encryption and Decryption." Information on digital signatures is available from "Digital Signatures."

Public-key cryptography is a set of well-established techniques and standards for protecting communications from eavesdropping, tampering, and impersonation attacks.

  • Encryption and decryption allow two communicating parties to disguise information they send to each other. The sender encrypts, or scrambles, information before sending it. The receiver decrypts, or unscrambles, the information after receiving it. While in transit, the encrypted information is unintelligible to an intruder.
  • Tamper detection allows the recipient of information to verify that it has not been modified in transit. Any attempt to modify data or substitute a false message for a legitimate one will be detected.
  • Authentication allows the recipient of information to determine its origin-that is, to confirm the sender's identity.
  • Nonrepudiation prevents the sender of information from claiming at a later date that the information was never sent.

The sections that follow introduce the concepts of public-key cryptography that underlie these capabilities.

Certificates and Authentication

A Certificate Identifies Someone or Something

A certificate is an electronic document used to identify an individual, a server, a company, or some other entity and to associate that identity with a public key. Like a driver's license, a passport, or other commonly used personal IDs, a certificate provides generally recognized proof of a person's identity. Public-key cryptography uses certificates to address the problem of impersonation.

To get a driver's license, you typically apply to a government agency, such as the Department of Motor Vehicles, which verifies your identity, your ability to drive, your address, and other information before issuing the license. To get a student ID, you apply to a school or college, which performs different checks (such as whether you have paid your tuition) before issuing the ID. To get a library card, you may need to provide only your name and a utility bill with your address on it.

Certificates work much the same way as any of these familiar forms of identification. Certificate authorities (CAs) are entities that validate identities and issue certificates. They can be either independent third parties or organizations running their own certificate-issuing server software (such as Red Hat Certificate System). The methods used to validate an identity vary depending on the policies of a given CA-just as the methods to validate other forms of identification vary depending on who is issuing the ID and the purpose for which it will be used. In general, before issuing a certificate, the CA must use its published verification procedures for that type of certificate to ensure that an entity requesting a certificate is in fact who it claims to be.

The certificate issued by the CA binds a particular public key to the name of the entity the certificate identifies (such as the name of an employee or a server). Certificates help prevent the use of fake public keys for impersonation. Only the public key certified by the certificate will work with the corresponding private key possessed by the entity identified by the certificate.

In addition to a public key, a certificate always includes the name of the entity it identifies, an expiration date, the name of the CA that issued the certificate, a serial number, and other information. Most importantly, a certificate always includes the digital signature of the issuing CA. The CA's digital signature allows the certificate to function as a "letter of introduction" for users who know and trust the CA but don't know the entity identified by the certificate.

For more information about the role of CAs, see "How CA Certificates Are Used to Establish Trust".

Authentication Confirms an Identity

Authentication is the process of confirming an identity. In the context of network interactions, authentication involves the confident identification of one party by another party. Authentication over networks can take many forms. Certificates are one way of supporting authentication.

Network interactions typically take place between a client, such as browser software running on a personal computer, and a server, such as the software and hardware used to host a Web site. Client authentication refers to the confident identification of a client by a server (that is, identification of the person assumed to be using the client software). Server authentication refers to the confident identification of a server by a client (that is, identification of the organization assumed to be responsible for the server at a particular network address).

Client and server authentication are not the only forms of authentication that certificates support. For example, the digital signature on an email message, combined with the certificate that identifies the sender, provide strong evidence that the person identified by that certificate did indeed send that message. Similarly, a digital signature on an HTML form, combined with a certificate that identifies the signer, can provide evidence, after the fact, that the person identified by that certificate did agree to the contents of the form. In addition to authentication, the digital signature in both cases ensures a degree of nonrepudiation-that is, a digital signature makes it difficult for the signer to claim later not to have sent the email or the form.

Client authentication is an essential element of network security within most intranets or extranets. The sections that follow contrast two forms of client authentication:

  • Password-Based Authentication. Almost all server software permits client authentication by means of a name and password. For example, a server might require a user to type a name and password before granting access to the server. The server maintains a list of names and passwords; if a particular name is on the list, and if the user types the correct password, the server grants access.
  • Certificate-Based Authentication. Client authentication based on certificates is part of the SSL protocol. The client digitally signs a randomly generated piece of data and sends both the certificate and the signed data across the network. The server uses techniques of public-key cryptography to validate the signature and confirm the validity of the certificate.

Password-Based Authentication

Figure 4 shows the basic steps involved in authenticating a client by means of a name and password. Figure 4 assumes the following:

  • The user has already decided to trust the server, either without authentication or on the basis of server authentication via SSL.
  • The user has requested a resource controlled by the server.
  • The server requires client authentication before permitting access to the requested resource.

Figure 4. Using a Password to Authenticate a Client to a Server

These are the steps shown in Figure 4:

  1. In response to an authentication request from the server, the client displays a dialog box requesting the user's name and password for that server. The user must supply a name and password separately for each new server the user wishes to use during a work session.
  2. The client sends the name and password across the network, either in the clear or over an encrypted SSL connection.
  3. The server looks up the name and password in its local password database and, if they match, accepts them as evidence authenticating the user's identity.
  4. The server determines whether the identified user is permitted to access the requested resource, and if so allows the client to access it.

With this arrangement, the user must supply a new password for each server, and the administrator must keep track of the name and password for each user, typically on separate servers.

Proper implementation does not store passwords in plaintext. Instead it concatenates the password with a random per-user value (so called "salt") and stores the hash value of the result along with the salt. This makes certain kinds of brute-force atacks more difficult.

As shown in the next section, one of the advantages of certificate-based authentication is that it can be used to replace the first three steps in Figure 4 with a mechanism that allows the user to supply just one password (which is not sent across the network) and allows the administrator to control user authentication centrally.

Certificate-Based Authentication

Figure 5 shows how client authentication works using certificates and the SSL protocol. To authenticate a user to a server, a client digitally signs a randomly generated piece of data and sends both the certificate and the signed data across the network. For the purposes of this discussion, the digital signature associated with some data can be thought of as evidence provided by the client to the server. The server authenticates the user's identity on the strength of this evidence.

Like Figure 4, Figure 5 assumes that the user has already decided to trust the server and has requested a resource, and that the server has requested client authentication in the process of evaluating whether to grant access to the requested resource.

Figure 5. Using a Certificate to Authenticate a Client to a Server

Unlike the process shown in Figure 4, the process shown in Figure 5 requires the use of SSL. Figure 5 also assumes that the client has a valid certificate that can be used to identify the client to the server. Certficate-based authentication is generally considered preferable to password-based authentication because it is based on wheat the user has (the private key) as well as what the user knows (the password that protects the private key). However, it's important to note that these two assumptions are true only if unauthorized personnel have not gained access to the user's machine or password, the password for the client software's private key database has been set, and the software is set up to request the password at reasonable frequent intervals.

Neither password-based authentication nor certificate-based authentication address security issues related to physical access to individual machines or passwords. Public-key cryptography can only verify that a private key used to sign some data corresponds to the public key in a certificate. It is the user's responsibility to protect a machine's physical security and to keep the private-key password secret.

These are the steps shown in Figure 5:

  1. The client software, such as Communicator, maintains a database of the private keys that correspond to the public keys published in any certificates issued for that client. The client asks for the password to this database the first time the client needs to access it during a given session-for example, the first time the user attempts to access an SSL-enabled server that requires certificate-based client authentication. After entering this password once, the user doesn't need to enter it again for the rest of the session, even when accessing other SSL-enabled servers.
  2. The client unlocks the private-key database, retrieves the private key for the user's certificate, and uses that private key to digitally sign some data that has been randomly generated for this purpose on the basis of input from both the client and the server. This data and the digital signature constitute "evidence" of the private key's validity. The digital signature can be created only with that private key and can be validated with the corresponding public key against the signed data, which is unique to the SSL session.
  3. The client sends both the user's certificate and the evidence (the randomly generated piece of data that has been digitally signed) across the network.
  4. The server uses the certificate and the evidence to authenticate the user's identity. (For a detailed discussion of the way this works, see "Introduction to SSL.")
  5. At this point the server may optionally perform other authentication tasks, such as checking that the certificate presented by the client is stored in the user's entry in an LDAP directory. The server then continues to evaluate whether the identified user is permitted to access the requested resource. This evaluation process can employ a variety of standard authorization mechanisms, potentially using additional information in an LDAP directory, company databases, and so on. If the result of the evaluation is positive, the server allows the client to access the requested resource.

As you can see by comparing Figure 5 to Figure 4, certificates replace the authentication portion of the interaction between the client and the server. Instead of requiring a user to send passwords across the network throughout the day, single sign-on requires the user to enter the private-key database password just once, without sending it across the network. For the rest of the session, the client presents the user's certificate to authenticate the user to each new server it encounters. Existing authorization mechanisms based on the authenticated user identity are not affected.

Uses for Certificates

Certificates have a purpose: to establish trust. Their usage varies depending on the kind of trust they are used to ensure. Some kinds of certificates are used to verify the identity of the presenter; others are used to verify that an object or item has not been tampered with.

SSL Protocol

The Secure Sockets Layer (SSL) protocol is a set of rules governing server authentication, client authentication, and encrypted communication between servers and clients. SSL is widely used on the Internet, especially for interactions that involve exchanging confidential information such as credit card numbers.

SSL requires a server SSL certificate, at a minimum. As part of the initial "handshake" process, the server presents its certificate to the client to authenticate the server's identity. The authentication process uses public-key encryption and digital signatures to confirm that the server is in fact the server it claims to be. Once the server has been authenticated, the client and server use techniques of symmetric-key encryption, which is very fast, to encrypt all the information they exchange for the remainder of the session and to detect any tampering that may have occurred.

Servers may optionally be configured to require client authentication as well as server authentication. In this case, after server authentication is successfully completed, the client must also present its certificate to the server to authenticate the client's identity before the encrypted SSL session can be established.

For an overview of client authentication over SSL and how it differs from password-based authentication, see "Authentication Confirms an Identity". For more detailed information about SSL, see "Introduction to SSL."

Signed and Encrypted Email

Some email programs support digitally signed and encrypted email using a widely accepted protocol known as Secure Multipurpose Internet Mail Extension (S/MIME). Using S/MIME to sign or encrypt email messages requires the sender of the message to have an S/MIME certificate.

An email message that includes a digital signature provides some assurance that it was in fact sent by the person whose name appears in the message header, thus providing authentication of the sender. If the digital signature cannot be validated by the email software on the receiving end, the user will be alerted.

The digital signature is unique to the message it accompanies. If the message received differs in any way from the message that was sent-even by the addition or deletion of a comma-the digital signature cannot be validated. Therefore, signed email also provides some assurance that the email has not been tampered with. As discussed at the beginning of this document, this kind of assurance is known as nonrepudiation. In other words, signed email makes it very difficult for the sender to deny having sent the message. This is important for many forms of business communication. (For information about the way digital signatures work, see "Digital Signatures".)

S/MIME also makes it possible to encrypt email messages. This is also important for some business users. However, using encryption for email requires careful planning. If the recipient of encrypted email messages loses his or her private key and does not have access to a backup copy of the key, for example, the encrypted messages can never be decrypted.

Single Sign-On

Network users are frequently required to remember multiple passwords for the various services they use. For example, a user might have to type a different password to log into the network, collect email, use directory services, use the corporate calendar program, and access various servers. Multiple passwords are an ongoing headache for both users and system administrators. Users have difficulty keeping track of different passwords, tend to choose poor ones, and tend to write them down in obvious places. Administrators must keep track of a separate password database on each server and deal with potential security problems related to the fact that passwords are sent over the network routinely and frequently.

Solving this problem requires some way for a user to log in once, using a single password, and get authenticated access to all network resources that user is authorized to use-without sending any passwords over the network. This capability is known as single sign-on.

Both client SSL certificates and S/MIME certificates can play a significant role in a comprehensive single sign-on solution. For example, one form of single sign-on relies on SSL client authentication (see "Certificate-Based Authentication"). A user can log in once, using a single password to the local client's private-key database, and get authenticated access to all SSL-enabled servers that user is authorized to use-without sending any passwords over the network. This approach simplifies access for users, because they don't need to enter passwords for each new server. It also simplifies network management, since administrators can control access by controlling lists of certificate authorities (CAs) rather than much longer lists of users and passwords.

In addition to using certificates, a complete single-sign on solution must address the need to interoperate with enterprise systems, such as the underlying operating system, that rely on passwords or other forms of authentication.

Object Signing

Communicator supports a set of tools and technologies called object signing. Object signing uses standard techniques of public-key cryptography to let users get reliable information about code they download in much the same way they can get reliable information about shrink-wrapped software.

Most importantly, object signing helps users and network administrators implement decisions about software distributed over intranets or the Internet-for example, whether to allow Java applets signed by a given entity to use specific computer capabilities on specific users' machines.

The "objects" signed with object signing technology can be applets or other Java code, JavaScript scripts, plug-ins, or any kind of file. The "signature" is a digital signature. Signed objects and their signatures are typically stored in a special file called a JAR file.

Software developers and others who wish to sign files using object-signing technology must first obtain an object-signing certificate.

Types of Certificates

Common types of certificates include the following:

  • Client SSL certificates. Used to identify clients to servers via SSL (client authentication). Typically, the identity of the client is assumed to be the same as the identity of a human being, such as an employee in an enterprise. See "Certificate-Based Authentication" for a description of the way client SSL certificates are used for client authentication. Client SSL certificates can also be used as part of a single sign-on solution.
Examples: A bank gives a customer a client SSL certificate that allows the bank's servers to identify that customer and authorize access to the customer's accounts. A company might give a new employee a client SSL certificate that allows the company's servers to identify that employee and authorize access to the company's servers.
  • Server SSL certificates. Used to identify servers to clients via SSL (server authentication). Server authentication may be used with or without client authentication. Server authentication is a requirement for an encrypted SSL session. For more information, see "SSL Protocol".
Example: Internet sites that engage in electronic commerce (commonly known as e-commerce) usually support certificate-based server authentication, at a minimum, to establish an encrypted SSL session and to assure customers that they are dealing with a web site identified with a particular company. The encrypted SSL session ensures that personal information sent over the network, such as credit card numbers, cannot easily be intercepted.
  • S/MIME certificates. Used for signed and encrypted email. As with client SSL certificates, the identity of the client is typically assumed to be the same as the identity of a human being, such as an employee in an enterprise. A single certificate may be used as both an S/MIME certificate and an SSL certificate (see "Signed and Encrypted Email"). S/MIME certificates can also be used as part of a single sign-on solution.
Examples: A company deploys combined S/MIME and SSL certificates solely for the purpose of authenticating employee identities, thus permitting signed email and client SSL authentication but not encrypted email. Another company issues S/MIME certificates solely for the purpose of both signing and encrypting email that deals with sensitive financial or legal matters.
  • Object-signing certificates. Used to identify signers of Java code, JavaScript scripts, or other signed files. For more information, see "Object Signing".
Example: A software company signs software distributed over the Internet to provide users with some assurance that the software is a legitimate product of that company. Using certificates and digital signatures in this manner can also make it possible for users to identify and control the kind of access downloaded software has to their computers.
Example: The CA certificates stored in Firefox determine what other certificates that copy of Firefox can authenticate. An administrator can implement some aspects of corporate security policies by controlling the CA certificates stored in each user's copy of Firefox.

Contents of a Certificate

The contents of certificates are organized according to the X.509 v3 certificate specification, which has been recommended by the International Telecommunications Union (ITU), an international standards body, since 1988.

Users don't usually need to be concerned about the exact contents of a certificate. However, system administrators working with certificates may need some familiarity with the information provided here.

Certificate Data Formats

Certificate requests and certificates can be created, stored, and installed in multiple formats: binary and text. All of these formats conform to X.509 standards. Three examples of binary formats are DER-encoded certificates, PKCS #7 certificate chains, and Netscape Certificate Sequence. Any of the binary formats can be imported in text form, which begins with the line:


Following this line is the certificate data, which can be in any of the binary formats described. This data should be base-64 encoded, as described by RFC 1113. The certificate information is followed by this line:


Distinguished Names

An X.509 v3 certificate binds a distinguished name (DN) to a public key. A DN is a series of name-value pairs, such as uid=doe, that uniquely identify an entity-that is, the certificate subject.

For example, this might be a typical DN for an employee of Example Corp:

uid=doe,,cn=John Doe,o=Example Corp.,c=US

The abbreviations before each equal sign in this example have these meanings:

  • uid: user ID
  • e: email address
  • cn: the user's common name
  • o: organization
  • c: country

DNs may include a variety of other name-value pairs. They are used to identify both certificate subjects and entries in directories that support the Lightweight Directory Access Protocol (LDAP).

The rules governing the construction of DNs can be quite complex and are beyond the scope of this document. For comprehensive information about DNs, see A String Representation of Distinguished Names] at the following URL:

A Typical Certificate

Every X.509 certificate consists of two sections:

  • The data section includes the following information:
    • The version number of the X.509 standard supported by the certificate.
    • The certificate's serial number. Every certificate issued by a CA has a serial number that is unique among the certificates issued by that CA.
    • Information about the user's public key, including the algorithm used and a representation of the key itself.
    • The DN of the CA that issued the certificate.
    • The period during which the certificate is valid (for example, between 1:00 p.m. on November 15, 1999 and 1:00 p.m. November 15, 2000)
    • The DN of the certificate subject (for example, in a client SSL certificate this would be the user's DN), also called the subject name.
    • Optional certificate extensions, which may provide additional data used by the client or server. For example, the certificate type extension indicates the type of certificate-that is, whether it is a client SSL certificate, a server SSL certificate, a certificate for signing email, and so on. Certificate extensions can also be used for a variety of other purposes.
  • The signature section includes the following information:
    • The cryptographic algorithm, or cipher, used by the issuing CA to create its own digital signature.
    • The CA's digital signature, obtained by hashing all of the data in the certificate together and encrypting it with the CA's private key.

Here are the data and signature sections of a certificate in human-readable format:

 Version: v3 (0x2)
 Serial Number: 3 (0x3)
 Signature Algorithm: PKCS #1 MD5 With RSA Encryption
 Issuer: OU=Ace Certificate Authority, O=Ace Industry, C=US
  Not Before: Fri Oct 17 18:36:25 1997
  Not  After: Sun Oct 17 18:36:25 1999
 Subject: CN=Jane Doe, OU=Finance, O=Ace Industry, C=US
 Subject Public Key Info:
  Algorithm: PKCS #1 RSA Encryption
  Public Key:
   Public Exponent: 65537 (0x10001)
   Identifier: Certificate Type
    Critical: no
    Certified Usage:
     SSL Client
   Identifier: Authority Key Identifier
    Critical: no
    Key Identifier:
   Algorithm: PKCS #1 MD5 With RSA Encryption

Here is the same certificate displayed in the 64-byte-encoded form interpreted by software:


How CA Certificates Are Used to Establish Trust

Certificate authorities (CAs) are entities that validate identities and issue certificates. They can be either independent third parties or organizations running their own certificate-issuing server software.

Any client or server software that supports certificates maintains a collection of trusted CA certificates. These CA certificates determine which other certificates the software can validate-in other words, which issuers of certificates the software can trust. In the simplest case, the software can validate only certificates issued by one of the CAs for which it has a certificate. It's also possible for a trusted CA certificate to be part of a chain of CA certificates, each issued by the CA above it in a certificate hierarchy.

The sections that follow explains how certificate hierarchies and certificate chains determine what certificates software can trust.

CA Hierarchies

In large organizations, it may be appropriate to delegate the responsibility for issuing certificates to several different certificate authorities. For example, the number of certificates required may be too large for a single CA to maintain; different organizational units may have different policy requirements; or it may be important for a CA to be physically located in the same geographic area as the people to whom it is issuing certificates.

It's possible to delegate certificate-issuing responsibilities to subordinate CAs. The X.509 standard includes a model for setting up a hierarchy of CAs like that shown in Figure 6.

Figure 6. Example of a Hierarchy of Certificate Authorities

In this model, the root CA is at the top of the hierarchy. The root CA's certificate is a self-signed certificate: that is, the certificate is digitally signed by the same entity-the root CA-that the certificate identifies. The CAs that are directly subordinate to the root CA have CA certificates signed by the root CA. CAs under the subordinate CAs in the hierarchy have their CA certificates signed by the higher-level subordinate CAs.

Organizations have a great deal of flexibility in terms of the way they set up their CA hierarchies. Figure 6 shows just one example; many other arrangements are possible.

Certificate Chains

CA hierarchies are reflected in certificate chains. A certificate chain is series of certificates issued by successive CAs. Figure 7 shows a certificate chain leading from a certificate that identifies some entity through two subordinate CA certificates to the CA certificate for the root CA (based on the CA hierarchy shown in Figure 6).

Figure 7. Example of a Certificate Chain

A certificate chain traces a path of certificates from a branch in the hierarchy to the root of the hierarchy. In a certificate chain, the following occur:

  • Each certificate is followed by the certificate of its issuer.
  • Each certificate contains the name (DN) of that certificate's issuer, which is the same as the subject name of the next certificate in the chain.

In Figure 7, the Engineering CA certificate contains the DN of the CA (that is, USA CA), that issued that certificate. USA CA's DN is also the subject name of the next certificate in the chain.

  • Each certificate is signed with the private key of its issuer. The signature can be verified with the public key in the issuer's certificate, which is the next certificate in the chain.

In Figure 7, the public key in the certificate for the USA CA can be used to verify the USA CA's digital signature on the certificate for the Engineering CA.

Verifying a Certificate Chain

Certificate chain verification is the process of making sure a given certificate chain is well-formed, valid, properly signed, and trustworthy. Red Hat software uses the following procedure for forming and verifying a certificate chain, starting with the certificate being presented for authentication:

  1. The certificate validity period is checked against the current time provided by the verifier's system clock.
  2. The issuer's certificate is located. The source can be either the verifier's local certificate database (on that client or server) or the certificate chain provided by the subject (for example, over an SSL connection).
  3. The certificate signature is verified using the public key in the issuer's certificate.
  4. If the issuer's certificate is trusted by the verifier in the verifier's certificate database, verification stops successfully here. Otherwise, the issuer's certificate is checked to make sure it contains the appropriate subordinate CA indication in the Red Hat certificate type extension, and chain verification returns to step 1 to start again, but with this new certificate. Figure 8 presents an example of this process.

Figure 8. Verifying a Certificate Chain All the Way to the Root CA

Figure 8 shows what happens when only Root CA is included in the verifier's local database. If a certificate for one of the intermediate CAs shown in Figure 8, such as Engineering CA, is found in the verifier's local database, verification stops with that certificate, as shown in Figure 9.

Figure 9. Verifying a Certificate Chain to an Intermediate CA

Expired validity dates, an invalid signature, or the absence of a certificate for the issuing CA at any point in the certificate chain causes authentication to fail. For example, Figure 10 shows how verification fails if neither the Root CA certificate nor any of the intermediate CA certificates are included in the verifier's local database.

Figure 10. A Certificate Chain That Can't Be Verified

For general information about the way digital signatures work, see "Digital Signatures". For a more detailed description of the signature verification process in the context of SSL client and server authentication, see "Introduction to SSL."

Managing Certificates

Certificates are used in many applications, from encrypting email to accessing websites. There are two major stages in the lifecycle of a certificate: the point when it is issued (issuance and enrollment) and the period when the certificates are no longer valid (renewal or revocation). There are also ways to manage the certificate during its lifecycle. Making information about the certificate available to other applications is publishing the certificate and then backing up the key pairs so that the certificate can be recovered if it is lost.

Issuing Certificates

The process for issuing a certificate depends on the certificate authority that issues it and the purpose for which it will be used. The process for issuing nondigital forms of identification varies in similar ways. For example, if you want to get a generic ID card (not a driver's license) from the Department of Motor Vehicles in California, the requirements are straightforward: you need to present some evidence of your identity, such as a utility bill with your address on it and a student identity card. If you want to get a regular driving license, you also need to take a test-a driving test when you first get the license, and a written test when you renew it. If you want to get a commercial license for an eighteen-wheeler, the requirements are much more stringent. If you live in some other state or country, the requirements for various kinds of licenses will differ.

Similarly, different CAs have different procedures for issuing different kinds of certificates. In some cases the only requirement may be your email address. In other cases, your UNIX or Windows login and password may be sufficient. At the other end of the scale, for certificates that identify people who can authorize large expenditures or make other sensitive decisions, the issuing process may require notarized documents, a background check, and a personal interview.

Depending on an organization's policies, the process of issuing certificates can range from being completely transparent for the user to requiring significant user participation and complex procedures. In general, processes for issuing certificates should be highly flexible, so organizations can tailor them to their changing needs.

Issuing certificates is one of several managements tasks that can be handled by separate Registration Authorities.

Certificates and the LDAP Directory

The Lightweight Directory Access Protocol (LDAP) for accessing directory services supports great flexibility in the management of certificates within an organization. System administrators can store much of the information required to manage certificates in an LDAP-compliant directory. For example, a CA can use information in a directory to prepopulate a certificate with a new employee's legal name and other information. The CA can leverage directory information in other ways to issue certificates one at a time or in bulk, using a range of different identification techniques depending on the security policies of a given organization. Other routine management tasks, such as key management and renewing and revoking certificates, can be partially or fully automated with the aid of the directory.

Information stored in the directory can also be used with certificates to control access to various network resources by different users or groups. Issuing certificates and other certificate management tasks can thus be an integral part of user and group management.

In general, high-performance directory services are an essential ingredient of any enterprise certificate management strategy.

Key Management

Before a certificate can be issued, the public key it contains and the corresponding private key must be generated. Sometimes it may be useful to issue a single person one certificate and key pair for signing operations, and another certificate and key pair for encryption operations. Separate signing and encryption certificates make it possible to keep the private signing key on the local machine only, thus providing maximum nonrepudiation, and to back up the private encryption key in some central location where it can be retrieved in case the user loses the original key or leaves the company.

Keys can be generated by client software or generated centrally by the CA and distributed to users via an LDAP directory. There are trade-offs involved in choosing between local and centralized key generation. For example, local key generation provides maximum nonrepudiation, but may involve more participation by the user in the issuing process. Flexible key management capabilities are essential for most organizations.

Key recovery, or the ability to retrieve backups of encryption keys under carefully defined conditions, can be a crucial part of certificate management (depending on how an organization uses certificates). Key recovery schemes usually involve an m of n mechanism: for example, m of n managers within an organization might have to agree, and each contribute a special code or key of their own, before a particular person's encryption key can be recovered. This kind of mechanism ensures that several authorized personnel must agree before an encryption key can be recovered.

Renewing and Revoking Certificates

Like a driver's license, a certificate specifies a period of time during which it is valid. Attempts to use a certificate for authentication before or after its validity period will fail. Therefore, mechanisms for managing certificate renewal are essential for any certificate management strategy. For example, an administrator may wish to be notified automatically when a certificate is about to expire, so that an appropriate renewal process can be completed in plenty of time without causing the certificate's subject any inconvenience. The renewal process may involve reusing the same public-private key pair or issuing a new one.

A driver's license can be suspended even if it has not expired-for example, as punishment for a serious driving offense. Similarly, it's sometimes necessary to revoke a certificate before it has expired-for example, if an employee leaves a company or moves to a new job within the company.

Certificate revocation can be handled in several different ways. For some organizations, it may be sufficient to set up servers so that the authentication process includes checking the directory for the presence of the certificate being presented. When an administrator revokes a certificate, the certificate can be automatically removed from the directory, and subsequent authentication attempts with that certificate will fail even though the certificate remains valid in every other respect. Another approach involves publishing a certificate revocation list (CRL)-that is, a list of revoked certificates-to the directory at regular intervals and checking the list as part of the authentication process. For some organizations, it may be preferable to check directly with the issuing CA each time a certificate is presented for authentication. This procedure is sometimes called real-time status checking.

Registration Authorities

Interactions between entities identified by certificates (sometimes called end entities) and CAs are an essential part of certificate management. These interactions include operations such as registration for certification, certificate retrieval, certificate renewal, certificate revocation, and key backup and recovery. In general, a CA must be able to authenticate the identities of end entities before responding to the requests. In addition, some requests need to be approved by authorized administrators or managers before being services.

As previously discussed, the means used by different CAs to verify an identity before issuing a certificate can vary widely, depending on the organization and the purpose for which the certificate will be used. To provide maximum operational flexibility, interactions with end entities can be separated from the other functions of a CA and handled by a separate service called a Registration Authority (RA).

An RA acts as a front end to a CA by receiving end entity requests, authenticating them, and forwarding them to the CA. After receiving a response from the CA, the RA notifies the end entity of the results. RAs can be helpful in scaling an PKI across different departments, geographical areas, or other operational units with varying policies and authentication requirements.

Original Document Information

Document Tags and Contributors

 Last updated by: kscarfone,