Red Hat SummitChapter 5. Planning the Certificate System
Each Red Hat Certificate System subsystem can be installed on the same server machine, installed on separate servers, or have multiple instances installed across an organization. Before installing any subsystem, it is important to plan the deployment out: what kind of PKI services do you need? What are the network requirements? What people need to access the Certificate System, what are their roles, and what are their physical locations? What kinds of certificates do you want to issue and what constraints or rules need to be set for them?
This chapter covers some basic questions for planning a Certificate System deployment. Many of these decisions are interrelated; one choice impacts another, like deciding whether to use smart cards determines whether to install the TPS and TKS subsystems.
5.1. Deciding on the required subsystems
The Certificate System subsystems cover different aspects of managing certificates. Planning which subsystems to install is one way of defining what PKI operations the deployment needs to perform.
Certificates, like software or equipment, have a lifecycle with defined stages. At its most basic, there are three steps:
- It is requested and issued.
- It is valid.
- It expires.
However, this simplified scenario does not cover a lot of common issues with certificates:
- What if an employee leaves the company before the certificate expires?
- When a CA signing certificate expires, all of the certificates issued and signed using that certificate also expire. So will the CA signing certificate be renewed, allowing its issued certificates to remain valid, or will it be reissued?
- What if an employee loses a smart card or leaves it at home. Will a replacement certificate be issued using the original certificates keys? Will the other certificates be suspended or revoked? Are temporary certificates allowed?
- When a certificate expires, will a new certificate be issued or will the original certificate be renewed?
This introduces three other considerations for managing certificates: revocation, renewal, and replacements.
Other considerations are the loads on the certificate authority. Are there a lot of issuance or renewal requests? Is there a lot of traffic from clients trying to validate whether certificates are valid? How are people requesting certificates supposed to authenticate their identity, and does that process slow down the issuance process?
5.1.1. Using a single Certificate Manager
The core of the Certificate System PKI is the Certificate Manager, a certificate authority. The CA receives certificate requests and issues all certificates.
Figure 5.1. CA only Certificate System
All of the basic processing for requests and issuing certificates can be handled by the Certificate Manager, and it is the only required subsystem. There can be a single Certificate Manager or many Certificate Managers, arranged in many different relationships, depending on the demands of the organization.
Along with issuing certificates, a Certificate Manager can also revoke certificates. One question for administrators is how to handle certificates if they are lost, compromised, or if the person or equipment for which they are issued leaves the company. Revoking a certificate invalidates it before its expiration date, and a list of revoked certificates is compiled and published by the issuing CA so that clients can verify the certificate status.
5.1.2. Planning for lost keys: key archival and recovery
One operation the CA cannot perform, though, is key archival and recovery. A very real scenario is that a user is going to lose his private key -for instance, the keys could be deleted from a browser database or a smart card could be left at home. Many common business operations use encrypted data, like encrypted email, and losing the keys which unlock that data means the data itself is lost. Depending on the policies in the company, there probably has to be a way to recover the keys in order to regenerate or reimport a replacement certificate, and both operations require the private key.
That requires a KRA, the subsystem which specially archives and retrieves keys.
Figure 5.2. CA and KRA
The Key Recovery Authority stores encryption keys (key archival) and can retrieve those keys so that the CA can reissue a certificate (key recovery). A KRA can store keys for any certificate generated by Certificate System, whether it is done for a regular certificate or when enrolling a smart card.
The key archival and recovery process is explained in more detail in Section 2.4.5, “Archiving, recovering, and rotating keys”.
The KRA is intended for archival and recovery of private encryption keys only. Therefore, end users must use a browser that supports dual-key generation to store their public-private key pairs.
5.1.3. Balancing certificate request processing
Another aspect of how the subsystems work together is load balancing. If a site has high traffic, then it is possible to simply install a lot of CAs, as clones of each other or in a flat hierarchy (where each CA is independent) or in a tree hierarchy (where some CAs are subordinate to other CAs); this is covered more in Section 5.2, “Defining the Certificate Authority hierarchy”.
5.1.4. Balancing client OCSP requests
If a certificate is within its validity period but needs be invalidated, it can be revoked. A Certificate Manager can publish lists of revoked certificates, so that when a client needs to verify that a certificate is still valid, it can check the list. These requests are online certificate status protocol requests, meaning that they have a specific request and response format. The Certificate Manager has a built-in OCSP responder so that it can verify OCSP requests by itself.
However, as with certificate request traffic, a site may have a significant number of client requests to verify certificate status. Example Corp. has a large web store, and each customer’s browser tries to verify the validity of their SSL/TLS certificates. Again, the CA can handle issuing the number of certificates, but the high request traffic affects its performance. In this case, Example Corp. uses the external OCSP Manager subsystem to verify certificate statuses, and the Certificate Manager only has to publish updated CRLs every so often.
Figure 5.3. CA and OCSP
5.1.5. Using smart cards
Smart cards usually require in-person enrollment and approval processes, since they use a physical medium to store key and certificate data. That means that multiple agents need to have access to the TPS and, possibly, multiple TPS subsystems in different offices or geographic locations.
The token management system is very scalable. Multiple TPSs can be configured to work with a single CA, TKS, or KRA instance, while multiple Enterprise Security Clients can communicate with a single TPS. As additional clients are installed, they can point back to the TPS instance without having to reconfigure the TPS; likewise, as TPSs are added, they can point to the same CA, TKS, and KRA instances without having to reconfigure those subsystems.
After installation, the TPS configuration can be edited to use additional CA, KRA, and TKS instances for failover support, so if the primary subsystem is unavailable, the TPS can switch to the next available system without interrupting its token services.
5.2. Defining the Certificate Authority hierarchy
The CA is the center of the PKI, so the relationship of CA systems, both to each other (CA hierarchy) and to other subsystems (security domain) is vital to planning a Certificate System PKI.
When there are multiple CAs in a PKI, the CAs are structured in a hierarchy or chain. The CA above another CA in a chain is called an root CA; a CA below another CA in the chain is called a subordinate CA. A CA can also be subordinate to a root outside of the Certificate System deployment; for example, a CA which functions as a root CA within the Certificate System deployment can be subordinate to a third-party CA.
A Certificate Manager (or CA) is subordinate to another CA because its CA signing certificate, the certificate that allows it to issue certificates, is issued by another CA. The CA that issued the subordinate CA signing certificate controls the CA through the contents of the CA signing certificate. The CA can constrain the subordinate CA through the kinds of certificates that it can issue, the extensions that it is allowed to include in certificates, the number of levels of subordinate CAs the subordinate CA can create, and the validity period of certificates it can issue, as well as the validity period of the subordinate CAs signing certificate.
Although a subordinate CA can create certificates that violate these constraints, a client authenticating a certificate that violates those constraints will not accept that certificate.
A self-signed root CA signs its own CA signing certificate and sets its own constraints as well as setting constraints on the subordinate CA signing certificates it issues.
A Certificate Manager can be configured as either a root CA or a subordinate CA. It is easiest to make the first CA installed a self-signed root, so that it is not necessary to apply to a third party and wait for the certificate to be issued. Before deploying the full PKI, however, consider whether to have a root CA, how many to have, and where both root and subordinate CAs will be located.
5.2.1. Subordination to a public CA
Chaining the Certificate System CA to a third-party public CA introduces the restrictions that public CAs place on the kinds of certificates the subordinate CA can issue and the nature of the certificate chain. For example, a CA that chains to a third-party CA might be restricted to issuing only Secure Multipurpose Internet Mail Extensions (S/MIME) and SSL/TLS client authentication certificates, but not SSL/TLS server certificates. There are other possible restrictions with using a public CA. This may not be acceptable for some PKI deployments.
One benefit of chaining to a public CA is that the third party is responsible for submitting the root CA certificate to a web browser or other client software. This can be a major advantage for an extranet with certificates that are accessed by different companies with browsers that cannot be controlled by the administrator. Creating a root CA in the CA hierarchy means that the local organization must get the root certificate into all the browsers which will use the certificates issued by the Certificate System. There are tools to do this within an intranet, but it can be difficult to accomplish with an extranet.
5.2.2. Subordination to a Certificate System CA
The Certificate System CA can function as a root CA, meaning that the server signs its own CA signing certificate as well as other CA signing certificates, creating an organization-specific CA hierarchy. The server can alternatively be configured as a subordinate CA, meaning the server’s CA signing key is signed by another CA in an existing CA hierarchy.
Setting up a Certificate System CA as the root CA means that the Certificate System administrator has control over all subordinate CAs by setting policies that control the contents of the CA signing certificates issued. A subordinate CA issues certificates by evaluating its own authentication and certificate profile configuration, without regard for the root CA’s configuration.
5.2.3. Linked CA
The Certificate System Certificate Manager can function as a linked CA, chaining up to many third-party or public CAs for validation; this provides cross-company trust, so applications can verify certificate chains outside the company certificate hierarchy. A Certificate Manager is chained to a third-party CA by requesting the Certificate Manager’s CA signing certificate from the third-party CA.
Related to this, the Certificate Manager also can issue cross-pair or cross-signed certificates. Cross-pair certificates create a trusted relationship between two separate CAs by issuing and storing cross-signed certificates between these two CAs. By using cross-signed certificate pairs, certificates issued outside the organization’s PKI can be trusted within the system.
These are also called bridge certificates, related to the Federal Bridge Certification Authority (FBCA) definition.
5.2.4. CA cloning
Instead of creating a hierarchy of root and subordinate CAs, it is possible to create multiple clones of a Certificate Manager and configure each clone to issue certificates within a range of serial numbers.
A cloned Certificate Manager uses the same CA signing key and certificate as another Certificate Manager, the master Certificate Manager.
If there is a chance that a subsystem will be cloned, then it is easiest to export its key pairs during the configuration process and save them to a secure location. The key pairs for the original Certificate Manager have to be available when the clone instance is configured, so that the clone can generate its certificates from the original Certificate Manager’s keys.
It is also possible to export the keys from the security databases at a later time, using the pk12util or the PKCS12Export commands.
Because clone CAs and original CAs use the same CA signing key and certificate to sign the certificates they issue, the issuer name in all the certificates is the same. Clone CAs and the original Certificate Managers issue certificates as if they are a single CA. These servers can be placed on different hosts for high availability failover support.
The advantage of cloning is that it distributes the Certificate Manager’s load across several processes or even several physical machines. For a CA with a high enrollment demand, the distribution gained from cloning allows more certificates to be signed and issued in a given time interval.
A cloned Certificate Manager has the same features, such as agent and end-entity gateway functions, of a regular Certificate Manager.
The serial numbers for certificates issued by clones are distributed dynamically. The databases for each clone and master are replicated, so all of the certificate requests and issued certificates, both, are also replicated. This ensures that there are no serial number conflicts while serial number ranges do not have to be manually assigned to the cloned Certificate Managers.
5.3. Planning security domains
A security domain is a registry of PKI services. PKI services, such as CAs, register information about themselves in these domains so users of PKI services can find other services by inspecting the registry. The security domain service in Certificate System manages both the registration of PKI services for Certificate System subsystems and a set of shared trust policies.
The registry provides a complete view of all PKI services provided by the subsystems within that domain. Each Certificate System subsystem must be either a host or a member of a security domain.
A CA subsystem is the only subsystem which can host a security domain. The security domain shares the CA internal database for privileged user and group information to determine which users can update the security domain, register new PKI services, and issue certificates.
A security domain is created during CA configuration, which automatically creates an entry in the security domain CA’s LDAP directory. Each entry contains all the important information about the domain. Every subsystem within the domain, including the CA registering the security domain, is recorded under the security domain container entry.
The URL to the CA uniquely identifies the security domain. The security domain is also given a friendly name, such as Example Corp Intranet PKI. All other subsystems - KRA, TPS, TKS, OCSP, and other CAs - must become members of the security domain by supplying the security domain URL when configuring the subsystem.
Each subsystem within the security domain shares the same trust policies and trusted roots which can be retrieved from different servers and browsers. The information available in the security domain is used during configuration of a new subsystem, which makes the configuration process streamlined and automated. For example, when a TPS needs to connect to a CA, it can consult the security domain to get a list of available CAs.
Each CA has its own LDAP entry. The security domain is an organizational group underneath that CA entry:
ou=Security Domain,dc=server.example.com-pki-ca
Then there is a list of each subsystem type beneath the security domain organizational group, with a special object class (pkiSecurityGroup) to identify the group type:
cn=KRAList,ou=Security Domain,o=pki-tomcat-CA objectClass: top objectClass: pkiSecurityGroup cn: KRAList
Each subsystem instance is then stored as a member of that group, with a special pkiSubsystem object class to identify the entry type:
dn: cn=kra.example.com:8443,cn=KRAList,ou=Security Domain,o=pki-tomcat-CA objectClass: top objectClass: pkiSubsystem cn: kra.example.com:8443 host: server.example.com UnSecurePort: 8080 SecurePort: 8443 SecureAdminPort: 8443 SecureAgentPort: 8443 SecureEEClientAuthPort: 8443 DomainManager: false Clone: false SubsystemName: KRA kra.example.com 8443
If a subsystem needs to contact another subsystem to perform an operation, it contacts the CA which hosts the security domain (by invoking a servlet which connects over the administrative port of the CA). The security domain CA then retrieves the information about the subsystem from its LDAP database, and returns that information to the requesting subsystem.
The subsystem authenticates to the security domain using a subsystem certificate.
Consider the following when planning the security domain:
- The CA hosting the security domain can be signed by an external authority.
- Multiple security domains can be set up within an organization. However, each subsystem can belong to only one security domain.
- Subsystems within a domain can be cloned. Cloning subsystem instances distributes the system load and provides failover points.
- The security domain streamlines configuration between the CA and KRA; the KRA can push its KRA connector information and transport certificates automatically to the CA instead of administrators having to manually copy the certificates over to the CA.
- The Certificate System security domain allows an offline CA to be set up. In this scenario, the offline root has its own security domain. All online subordinate CAs belong to a different security domain.
- The security domain streamlines configuration between the CA and OCSP. The OCSP can push its information to the CA for the CA to set up OCSP publishing and also retrieve the CA certificate chain from the CA and store it in the internal database.
5.4. Determining the requirements for subsystem certificates
The CA configuration determines many of the characteristics of the certificates which it issues, regardless of the actual type of certificate being issued. Constraints on the CA’s own validity period, distinguished name, and allowed encryption algorithms impact the same characteristics in their issued certificates. Additionally, the Certificate Managers have predefined profiles that set rules for different kinds of certificates that they issue, and additional profiles can be added or modified. These profile configurations also impact issued certificates.
5.4.1. Determining which certificates to install
When a Certificate System subsystem is first installed and configured, the certificates necessary to access and administer it are automatically created. These include an agent’s certificate, server certificate, and subsystem-specific certificates. These initial certificates are shown in Table 5.1, “Initial subsystem certificates”.
| Subsystem | Certificates |
|---|---|
| Certificate Manager |
|
| OCSP |
|
| KRA |
|
| TKS |
|
| TPS |
|
There are some cautionary considerations about replacing existing subsystem certificates.
Generating new key pairs when creating a new self-signed CA certificate for a root CA will invalidate all certificates issued under the previous CA certificate.
This means none of the certificates issued or signed by the CA using its old key will work; subordinate Certificate Managers, KRAs, OCSPs, TKSs, and TPSs will no longer function, and agents can no longer to access agent interfaces.
This same situation occurs if a subordinate CA’s CA certificate is replaced by one with a new key pair; all certificates issued by that CA are invalidated and will no longer work.
Instead of creating new certificates from new key pairs, consider renewing the existing CA signing certificate.
- If the CA is configured to publish to the OCSP and it has a new CA signing certificate or a new CRL signing certificate, the CA must be identified again to the OCSP.
-
If a new transport certificate is created for the KRA, the KRA information must be updated in the CA’s configuration file,
CS.cfg. The existing transport certificate must be replaced with the new one in theca.connector.KRA.transportCertparameter. - If a CA is cloned, then when creating a new SSL/TLS server certificate for the master Certificate Manager, the clone CAs' certificate databases all need updated with the new SSL/TLS server certificate.
- If the Certificate Manager is configured to publish certificates and CRLs to an LDAP directory and uses the SSL/TLS server certificate for SSL/TLS client authentication, then the new SSL/TLS server certificate must be requested with the appropriate extensions. After installing the certificate, the publishing directory must be configured to use the new server certificate.
- Any number of SSL/TLS server certificates can be issued for a subsystem instance, but it really only needs one SSL/TLS certificate. This certificate can be renewed or replaced as many times as necessary.
5.4.2. Planning the CA distinguished name
The core elements of a CA are a signing unit and the Certificate Manager identity. The signing unit digitally signs certificates requested by end entities. A Certificate Manager must have its own distinguished name (DN), which is listed in every certificate it issues.
Like any other certificate, a CA certificate binds a DN to a public key. A DN is a series of name-value pairs that in combination uniquely identify an entity. For example, the following DN identifies a Certificate Manager for the Engineering department of a corporation named Example Corporation:
cn=demoCA, o=Example Corporation, ou=Engineering, c=US
Many combinations of name-value pairs are possible for the Certificate Manager’s DN. The DN must be unique and readily identifiable, since any end entity can examine it.
5.4.3. Setting the CA signing certificate validity period
Every certificate, including a Certificate Manager signing certificate, must have a validity period. The Certificate System does not restrict the validity period that can be specified. Set as long a validity period as possible, depending on the requirements for certificate renewal, the place of the CA in the certificate hierarchy, and the requirements of any public CAs that are included in the PKI.
A Certificate Manager cannot issue a certificate that has a validity period longer than the validity period of its CA signing certificate. If a request is made for a period longer than the CA certificate’s validity period, the requested validity date is ignored and the CA signing certificate validity period is used.
5.4.4. Choosing the signing key type and length
A signing key is used by a subsystem to verify and "seal" something. CAs use a CA signing certificate to sign certificates or CRLs that it issues; OCSPs use signing certificates to verify their responses to certificate status requests; all subsystems use log file signing certificates to sign their audit logs.
The signing key must be cryptographically strong to provide protection and security for its signing operations. The following signing algorithms are considered secure:
- SHA256withRSA
- SHA512withRSA
- SHA256withEC
- SHA512withEC
Certificate System includes native ECC support. It is also possible to load and use a third-party PKCS #11 module with ECC-enabled. This is covered in Section 7.2, “Installing RHCS using the pkispawn utility”.
Along with a key type, each key has a specific bit length. Longer keys are considered cryptographically stronger than shorter keys. However, longer keys require more time for signing operations.
The default RSA key length in the configuration wizard is 2048 bits; for certificates that provide access to highly sensitive data or services, consider increasing the length to 4096 bits. ECC keys are much stronger than RSA keys, so the recommended length for ECC keys is 256 bits, which is equivalent in strength to a 2048-bit RSA key.
5.4.5. Using certificate extensions
An X.509 v3 certificate contains an extension field that permits any number of additional fields to be added to the certificate. Certificate extensions provide a way of adding information such as alternative subject names and usage restrictions to certificates. Older Netscape servers, such as Red Hat Directory Server and Red Hat Certificate System, require Netscape-specific extensions because they were developed before PKIX part 1 standards were defined.
The X.509 v1 certificate specification was originally designed to bind public keys to names in an X.500 directory. As certificates began to be used on the Internet and extranets and directory lookups could not always be performed, problem areas emerged that were not covered by the original specification.
- Trust. The X.500 specification establishes trust by means of a strict directory hierarchy. By contrast, Internet and extranet deployments frequently involve distributed trust models that do not conform to the hierarchical X.500 approach.
- Certificate use. Some organizations restrict how certificates are used. For example, some certificates may be restricted to client authentication only.
- Multiple certificates. It is not uncommon for certificate users to possess multiple certificates with identical subject names but different key material. In this case, it is necessary to identify which key and certificate should be used for what purpose.
- Alternate names. For some purposes, it is useful to have alternative subject names that are also bound to the public key in the certificate.
- Additional attributes. Some organizations store additional information in certificates, such as when it is not possible to look up information in a directory.
- Relationship with CA. When certificate chaining involves intermediate CAs, it is useful to have information about the relationships among CAs embedded in their certificates.
- CRL checking. Since it is not always possible to check a certificate’s revocation status against a directory or with the original certificate authority, it is useful for certificates to include information about where to check CRLs.
The X.509 v3 specification addressed these issues by altering the certificate format to include additional information within a certificate by defining a general format for certificate extensions and specifying extensions that can be included in the certificate. The extensions defined for X.509 v3 certificates enable additional attributes to be associated with users or public keys and manage the certification hierarchy. The Internet X.509 Public Key Infrastructure Certificate and CRL Profile recommends a set of extensions to use for Internet certificates and standard locations for certificate or CA information. These extensions are called standard extensions.
The X.509 v3 standard for certificates allows organizations to define custom extensions and include them in certificates. These extensions are called private, proprietary, or custom extensions, and they carry information unique to an organization or business. Applications may not able to validate certificates that contain private critical extensions, so it not recommended that these be used in wide-spread situations.
The X.500 and X.509 specifications are controlled by the International Telecommunication Union (ITU), an international organization that primarily serves large telecommunication companies, government organizations, and other entities concerned with the international telecommunications network. The Internet Engineering Task Force (IETF), which controls many of the standards that underlie the Internet, is currently developing public-key infrastructure X.509 (PKIX) standards. These proposed standards further refine the X.509 v3 approach to extensions for use on the Internet. The recommendations for certificates and CRLs have reached proposed standard status and are in a document referred to as PKIX Part 1.
Two other standards, Abstract Syntax Notation One (ASN.1) and Distinguished Encoding Rules (DER), are used with Certificate System and certificates in general. These are specified in the CCITT Recommendations X.208 and X.209. For a quick summary of ASN.1 and DER, see A Layman’s Guide to a Subset of ASN.1, BER, and DER, which is available at RSA Laboratories' web site, http://www.rsa.com.
5.4.5.1. Structure of certificate extensions
In RFC 3280, an X.509 certificate extension is defined as follows:
Extension ::= SEQUENCE {
extnID OBJECT IDENTIFIER,
critical BOOLEAN DEFAULT FALSE,
extnValue OCTET STRING }The means a certificate extension consists of the following:
-
The object identifier (OID) for the extension. This identifier uniquely identifies the extension. It also determines the ASN.1 type of value in the value field and how the value is interpreted. When an extension appears in a certificate, the OID appears as the extension ID field (
extnID) and the corresponding ASN.1 encoded structure appears as the value of the octet string (extnValue). A flag or Boolean field called
critical.The value, which can be either
trueorfalse, assigned to this field indicates whether the extension is critical or noncritical to the certificate. If the extension is critical and the certificate is sent to an application that does not understand the extension based on the extension’s ID, the application must reject the certificate. If the extension is not critical and the certificate is sent to an application that does not understand the extension based on the extension’s ID, the application can ignore the extension and accept the certificate.- An octet string containing the DER encoding of the value of the extension.
Typically, the application receiving the certificate checks the extension ID to determine if it can recognize the ID. If it can, it uses the extension ID to determine the type of value used.
Some of the standard extensions defined in the X.509 v3 standard include the following:
- Authority Key Identifier extension, which identifies the CA’s public key, the key used to sign the certificate.
- Subject Key Identifier extension, which identifies the subject’s public key, the key being certified.
Not all applications support certificates with version 3 extensions. Applications that do support these extensions may not be able to interpret some or all of these specific extensions.
5.4.6. Using and customizing certificate profiles
Certificates have different types and different applications. They can be used to establish a single sign-on environment for a corporate network, to set up VPNs, to encrypt email, or to authenticate to a website. The requirements for all of these certificates can be different, just as there may also be different requirements for the same type of certificate for different kinds of users. These certificate characteristics are set in certificate profiles. The Certificate Manager defines a set of certificate profiles that it uses as enrollment forms when users or machines request certificates.
Certificate profiles
A certificate profile defines everything associated with issuing a particular type of certificate, including the authentication method, the certificate content (defaults), constraints for the values of the content, and the contents of the input and output for the certificate profile. Enrollment requests are submitted to a certificate profile and are then subject to the defaults and constraints set in that certificate profile. These constraints are in place whether the request is submitted through the input form associated with the certificate profile or through other means. The certificate that is issued from a certificate profile request contains the content required by the defaults with the information required by the default parameters. The constraints provide rules for what content is allowed in the certificate.
For example, a certificate profile for user certificates defines all aspects of that certificate, including the validity period of the certificate. The default validity period can be set to two years, and a constraint can be set on the profile that the validity period for certificates requested through this certificate profile cannot exceed two years. When a user requests a certificate using the input form associated with this certificate profile, the issued certificate contains the information specified in the defaults and will be valid for two years. If the user submits a pre-formatted request for a certificate with a validity period of four years, the request is rejected since the constraints allow a maximum of two years validity period for this type of certificate.
A set of certificate profiles have been predefined for the most common certificates issued. These certificate profiles define defaults and constraints, associate the authentication method, and define the needed inputs and outputs for the certificate profile.
Modifying the certificate profile parameters
The parameters of the default certificate profiles can be modified; this includes the authentication method, the defaults, the constraints used in each profile, the values assigned to any of the parameters in a profile, the input, and the output. It is also possible to create new certificate profiles for other types of certificates or for creating more than one certificate profile for a certificate type. There can be multiple certificate profiles for a particular type of certificate to issue the same type of certificate with a different authentication method or different definitions for the defaults and constraints. For example, there can be two certificate profiles for enrollment of SSL/TLS server certificates where one certificate profile issues certificates with a validity period of six months and another certificate profile issues certificates with a validity period of two years.
An input sets a text field in the enrollment form and what kind of information needs gathered from the end entity; this includes setting the text area for a certificate request to be pasted, which allows a request to be created outside the input form with any of the request information required. The input values are set as values in the certificate. The default inputs are not configurable in the Certificate System.
An output specifies how the response page to a successful enrollment is presented. It usually displays the certificate in a user-readable format. The default output shows a printable version of the resultant certificate; other outputs set the type of information generated at the end of the enrollment, such as PKCS #7.
Policy sets are sets of constraints and default extensions attached to every certificate processed through the profile. The extensions define certificate content such as validity periods and subject name requirements. A profile handles one certificate request, but a single request can contain information for multiple certificates. A PKCS#10 request contains a single public key. One CRMF request can contain multiple public keys, meaning multiple certificate requests. A profile may contain multiple sets of policies, with each set specifying how to handle one certificate request within a CRMF request.
Certificate Profile Administration
An administrator sets up a certificate profile by associating an existing authentication plug-in, or method, with the certificate profile; enabling and configuring defaults and constraints; and defining inputs and outputs. The administrator can use the existing certificate profiles, modify the existing certificate profiles, create new certificate profiles, and delete any certificate profile that will not be used in this PKI.
Once a certificate profile is set up, it appears on the Manage Certificate Profiles page of the agent services page where an agent can approve, and thus enable, a certificate profile. Once the certificate profile is enabled, it appears on the Certificate Profile tab of the end-entities page where end entities can enroll for a certificate using the certificate profile.
The certificate profile enrollment page in the end-entities interface contains links to each certificate profile that has been enabled by the agents. When an end entity selects one of those links, an enrollment page appears containing an enrollment form specific to that certificate profile. The enrollment page is dynamically generated from the inputs defined for the profile. If an authentication plug-in is configured, additional fields may be added to authenticate the user.
When an end entity submits a certificate profile request that is associated with an agent-approved (manual) enrollment, an enrollment where no authentication plug-in is configured, the certificate request is queued in the agent services interface. The agent can change some aspects of the enrollment, request, validate it, cancel it, reject it, update it, or approve it. The agent is able to update the request without submitting it or validate that the request adheres to the profile’s defaults and constraints. This validation procedure is only for verification and does not result in the request being submitted. The agent is bound by the constraints set; they cannot change the request in such a way that a constraint is violated. The signed approval is immediately processed, and a certificate is issued.
When a certificate profile is associated with an authentication method, the request is approved immediately and generates a certificate automatically if the user successfully authenticates, all the information required is provided, and the request does not violate any of the constraints set up for the certificate profile. There are profile policies which allow user-supplied settings like subject names or validity periods. The certificate profile framework can also preserve user-defined content set in the original certificate request in the issued certificate.
The issued certificate contains the content defined in the defaults for this certificate profile, such as the extensions and validity period for the certificate. The content of the certificate is constrained by the constraints set for each default. Multiple policies (defaults and constraints) can be set for one profile, distinguishing each set by using the same value in the policy set ID. This is particularly useful for dealing with dual keys enrollment where encryption keys and signing keys are submitted to the same profile. The server evaluates each set with each request it receives. When a single certificate is issued, one set is evaluated, and any other sets are ignored. When dual-key pairs are issued, the first set is evaluated with the first certificate request, and the second set is evaluated with the second certificate request. There is no need for more than one set for issuing a single certificate or more than two sets for issuing dual-key pairs.
Guidelines for customizing certificate profiles
Tailor the profiles for the organization to the real needs and anticipated certificate types used by the organization:
- Decide which certificate profiles are needed in the PKI. There should be at least one profile for each type of certificate issued. There can be more than one certificate profile for each type of certificate to set different authentication methods or different defaults and constraints for a particular type of certificate type. Any certificate profile available in the administrative interface can be approved by an agent and then used by an end entity to enroll.
- Delete any certificate profiles that will not be used.
Modify the existing certificate profiles for specific characteristics for the company’s certificates.
- Change the defaults set up in the certificate profile, the values of the parameters set in the defaults, or the constraints that control the certificate content.
- Change the constraints set up by changing the value of the parameters.
- Change the authentication method.
- Change the inputs by adding or deleting inputs in the certificate profile, which control the fields on the input page.
- Add or delete the output.
5.4.6.1. Adding SAN extensions to the SSL server certificate
Certificate System enables adding Subject Alternative Name (SAN) extensions to the SSL server certificate during the installation of non-root CA or other Certificate System instances. To do so, follow the instructions in the /usr/share/pki/ca/profiles/ca/caInternalAuthServerCert.cfg file and add the following parameters to the configuration file supplied to the pkispawn utility:
pki_san_inject-
Set the value of this parameter to
True. pki_san_for_server_cert- Provide a list of the required SAN extensions separated by commas (,).
For example:
pki_san_inject=True pki_san_for_server_cert=intca01.example.com,intca02.example.com,intca.example.com
5.4.7. Planning authentication methods
As implied in Section 5.4.6, “Using and customizing certificate profiles”, authentication for the certificate process means the way that a user or entity requesting a certificate proves that they are who they say they are. There are three ways that the Certificate System can authenticate an entity:
- In agent-approved enrollment, end-entity requests are sent to an agent for approval. The agent approves the certificate request.
- In automatic enrollment, end-entity requests are authenticated using a plug-in, and then the certificate request is processed; an agent is not involved in the enrollment process.
- In CMC enrollment, a third party application can create a request that is signed by an agent and then automatically processed.
A Certificate Manager is initially configured for agent-approved enrollment and for CMC authentication. Automated enrollment is enabled by configuring one of the authentication plug-in modules. More than one authentication method can be configured in a single instance of a subsystem. The HTML registration pages contain hidden values specifying the method used. With certificate profiles, the end-entity enrollment pages are dynamically-generated for each enabled profile. The authentication method associated with this certificate profile is specified in the dynamically-generated enrollment page.
The authentication process is simple.
- An end entity submits a request for enrollment. The form used to submit the request identifies the method of authentication and enrollment. All HTML forms are dynamically-generated by the profiles, which automatically associate the appropriate authentication method with the form.
- If the authentication method is an agent-approved enrollment, the request is sent to the request queue of the CA agent. If the automated notification for a request in queue is set, an email is sent to the appropriate agent that a new request has been received. The agent can modify the request as allowed for that form and the profile constraints. Once approved, the request must pass the certificate profiles set for the Certificate Manager, and then the certificate is issued. When the certificate is issued, it is stored in the internal database and can be retrieved by the end entity from the end-entities page by serial number or by request ID.
- If the authentication method is automated, the end entity submits the request along with required information to authenticate the user, such as an LDAP user name and password. When the user is successfully authenticated, the request is processed without being sent to an agent’s queue. If the request passes the certificate profile configuration of the Certificate Manager, the certificate is issued and stored in the internal database. It is delivered to the end entity immediately through the HTML forms.
The requirements for how a certificate request is authenticated can have a direct impact on the necessary subsystems and profile settings. For example, if an agent-approved enrollment requires that an agent meet the requester in person and verify their identity through supported documentation, the authentication process can be time-intensive, as well as constrained by the physical availability of both the agent and the requester.
5.4.8. Publishing certificates and CRLs
A CA can publish both certificates and CRLs. Certificates can be published to a plain file or to an LDAP directory; CRLs can be published to file or an LDAP directory, as well, and can also be published to an OCSP responder to handle certificate verification.
Configuring publishing is fairly straightforward and is easily adjusted. For continuity and accessibility, though, it is good to plan out where certificates and CRLs need to be published and what clients need to be able to access them.
Publishing to an LDAP directory requires special configuration in the directory for publishing to work:
- If certificates are published to the directory, than every user or server to which a certificate is issued must have a corresponding entry in the LDAP directory.
- If CRLs are published to the directory, than they must be published to an entry for the CA which issued them.
- For SSL/TLS, the directory service has to be configured in SSL/TLS and, optionally, be configured to allow the Certificate Manager to use certificate-based authentication.
- The directory administrator should configure appropriate access control rules to control DN (entry name) and password based authentication to the LDAP directory.
5.4.9. Renewing or reissuing CA signing certificates
When a CA signing certificate expires, all certificates signed with the CA’s corresponding signing key become invalid. End entities use information in the CA certificate to verify the certificate’s authenticity. If the CA certificate itself has expired, applications cannot chain the certificate to a trusted CA.
There are two ways of resolving CA certificate expiration:
- Renewing a CA certificate involves issuing a new CA certificate with the same subject name and public and private key material as the old CA certificate, but with an extended validity period. As long as the new CA certificate is distributed to all users before the old CA certificate expires, renewing the certificate allows certificates issued under the old CA certificate to continue working for the full duration of their validity periods.
- Reissuing a CA certificate involves issuing a new CA certificate with a new name, public and private key material, and validity period. This avoids some problems associated with renewing a CA certificate, but it requires more work for both administrators and users to implement. All certificates issued by the old CA, including those that have not yet expired, must be renewed by the new CA.
There are problems and advantages with either renewing or reissuing a CA certificate. Begin planning the CA certificate renewal or re-issuance before installing any Certificate Managers, and consider the ramifications the planned procedures may have for extensions, policies, and other aspects of the PKI deployment.
Correct use of extensions, for example the authorityKeyIdentifier extension, can affect the transition from an old CA certificate to a new one.
5.5. Planning for network and physical security
When deploying any Certificate System subsystem, the physical and network security of the subsystem instance has to be considered because of the sensitivity of the data generated and stored by the subsystems.
5.5.1. Considering firewalls
There are two considerations about using firewalls with Certificate System subsystems:
- Protecting sensitive subsystems from unauthorized access
- Allowing appropriate access to other subsystems and clients outside of the firewall
The CA, KRA, and TKS are always placed inside a firewall because they contain critical information that can cause devastating security consequences if they are compromised.
The TPS and OCSP can be placed outside the firewall. Likewise, other services and clients used by the Certificate System can be on a different machine outside the firewall. In that case, the local networks have to be configured to allow access between the subsystems behind the firewall and the services outside it.
The LDAP database can be on a different server, even on a different network, than the subsystem which uses it. In this case, all LDAP ports (389 for LDAP and 636 for LDAPS, by default) need to be open in the firewall to allow traffic to the directory service. Without access to the LDAP database, all subsystem operations can fail.
As part of configuring the firewalls, if iptables is enabled, then it must have configured policies to allow communication over the appropriate Certificate System ports. Configuring iptables is described in the Using and configuring firewalld guide.
5.5.2. Considering physical security and location
Because of the sensitive data they hold, consider keeping the CA, KRA, and TKS in a secure facility with adequate physical access restrictions. Just as network access to the systems needs to be restricted, the physical access also needs to be restricted.
Along with finding a secure location, consider the proximity to the subsystem agents and administrators. Key recovery, for example, can require multiple agents to give approval; if these agents are spread out over a large geographical area, then the time differences may negatively impact the ability to retrieve keys. Plan the physical locations of the subsystems, then according to the locations of the agents and administrators who will manage the subsystem.
5.5.3. Planning ports
Each Certificate System server uses up to four ports:
- A non-secure HTTP port for end entity services that do not require authentication
- A secure HTTP port for end entity services, agent services, administrative console, and admin services that require authentication
- A Tomcat Server Management port
- A Tomcat AJP Connector Port
All of the service pages and interfaces described in the Red Hat Certificate System User Interfaces chapter in the Red Hat Certificate System Administration Guide are connected to using the instance’s URL and the corresponding port number. For example:
https://server.example.com:8443/ca/ee/ca
To access the admin console, the URL specifies the admin port:
pkiconsole https://server.example.com:8443/ca
All agent and admin functions require SSL/TLS client authentication. For requests from end entities, the Certificate System listens on both the SSL/TLS (encrypted) port and non-SSL/TLS ports.
The ports are defined in the server.xml file. If a port is not used, it can be disabled in that file. For example:
<Service name="Catalina">
<!--Connector port="8080" ... /--> unused standard port
<Connector port="8443" ... />Whenever a new instance in installed, make sure that the new ports are unique on the host system.
To verify that a port is available for use, check the appropriate file for the operating system. Port numbers for network-accessible services are usually maintained in a file named services. On Red Hat Enterprise Linux, it is also helpful to confirm that a port is not assigned by SELinux, by running the command semanage port -l to list all ports which currently have an SELinux context.
When a new subsystem instance is created, any number between 1 and 65535 can be specified as the secure port number.
5.6. Tokens for storing Certificate System subsystem keys and certificates
A token is a hardware or software device that performs cryptographic functions and stores public-key certificates, cryptographic keys, and other data. The Certificate System defines two types of tokens, internal and external, for storing key pairs and certificates that belong to the Certificate System subsystems.
An internal (software) token is a pair of files, usually called the certificate database (cert9.db) and key database (key4.db), that the Certificate System uses to generate and store its key pairs and certificates. The Certificate System automatically generates these files in the filesystem of its host machine when first using the internal token. These files are created during the Certificate System subsystem configuration if the internal token was selected for key-pair generation.
These security databases are located in the /var/lib/pki/ instance_name/alias directory.
An external token refers to an external hardware device, such as a smart card or hardware security module (HSM), that the Certificate System uses to generate and store its key pairs and certificates. The Certificate System supports any hardware tokens that are compliant with PKCS #11.
PKCS #11 is a standard set of APIs and shared libraries which isolate an application from the details of the cryptographic device. This enables the application to provide a unified interface for PKCS #11-compliant cryptographic devices.
The PKCS #11 module implemented in the Certificate System supports cryptographic devices supplied by many different manufacturers. This module allows the Certificate System to plug in shared libraries supplied by manufacturers of external encryption devices and use them for generating and storing keys and certificates for the Certificate System managers.
Consider using external tokens for generating and storing the key pairs and certificates used by Certificate System. These devices are another security measure to safeguard private keys because hardware tokens are sometimes considered more secure than software tokens.
Before using external tokens, plan how the external token is going to be used with the subsystem:
- All system keys for a subsystem must be generated on the same token.
- The subsystem must be installed in an empty HSM slot. If the HSM slot has previously been used to store other keys, then use the HSM vendor’s utilities to delete the contents of the slot. The Certificate System has to be able to create certificates and keys on the slot with default nicknames. If not properly cleaned up, the names of these objects may collide with previous instances.
The Certificate System can also use hardware cryptographic accelerators with external tokens. Many of the accelerators provide the following security features:
- Fast SSL/TLS connections. Speed is important to accommodate a high number of simultaneous enrollment or service requests.
- Hardware protection of private keys. These devices behave like smart cards by not allowing private keys to be copied or removed from the hardware token. This is important as a precaution against key theft from an active attack of an online Certificate Manager.
The Certificate System supports the nCipher nShield hardware security module (HSM), by default. Certificate System-supported HSMs are automatically added to the pkcs11.txt database with modutil during the pre-configuration stage of the installation, if the PKCS #11 library modules are in the default installation paths.
During configuration, the Security Modules panel displays the supported modules, along with the NSS internal software PKCS#11 module. All supported modules that are detected show a status of Found and is individually marked as either Logged in or Not logged in. If a token is found but not logged in, it is possible to log in using the under Operations. If the administrator can log into a token successfully, the password is stored in a configuration file. At the next start or restart of the Certificate System instance, the passwords in the password store are used to attempt a login for each corresponding token.
Administrators are allowed to select any of the tokens that are logged in as the default token, which is used to generate system keys.
5.7. A Checklist for planning the PKI
- How many security domains will be created, and what subsystem instances will be placed in each domain?
- A subsystem can only communicate with another subsystem if they have a trusted relationship. Because the subsystems within a security domain have automatic trusted relationships with each other, it is important what domain a subsystem joins. Security domains can have different certificate issuing policies, different kinds of subsystems within them, or a different Directory Server database. Map out where (both on the physical machine and in relation to each other) each subsystem belongs, and assign it to the security domain accordingly.
- What ports should be assigned for each subsystem? Is it necessary to have a single SSL/TLS port, or is it better to have port separation for extra security?
- The most secure solution is to use separate ports for each SSL/TLS interface. However, the feasibility of implementing multiple ports may depends on your network setup, firewall rules, and other network conditions.
- What subsystems should be placed behind firewalls? What clients or other subsystems need to access those firewall-protected subsystems and how will access be granted? Is firewall access allowed for the LDAP database?
The location to install a subsystem depends on the network design. The OCSP subsystem is specifically designed to operate outside a firewall for user convenience, while the CA, KRA, and TPS should all be secured behind a firewall for increased security.
When deciding the location of the subsystems, it is critical to plan what other subsystems or services that server needs to access (including the internal database, a CA, or external users) and look at firewall, subnet, and VPN configuration.
- What subsystems need to be physically secured? How will access be granted, and who will be granted access?
- The CA, TKS, and KRA all store extremely sensitive key and certificate information. For some deployments, it may be necessary to limit physical access to the machines these subsystems run on. In that case, both the subsystems and their host machines must be included in the larger infrastructure inventory and security design.
- What is the physical location of all agents and administrators? What is the physical location of the subsystems? How will administrators or agents access the subsystem services in a timely-manner? Is it necessary to have subsystems in each geographical location or time zone?
- The physical location of Certificate System users may impact things like request approval and token enrollment, as well as system performance because of network lag time. The importance of response time for processing certificate operations should be taken into account when deciding where and how many subsystems to install.
- How many subsystems do you need to install?
The primary consideration is the expected load for each subsystem and then, secondarily, geographical or departmental divisions. Subsystems with fairly low loads (like the KRA or TKS) may only require a single instance for the entire PKI. Systems with high load (the CA) or which may benefit from local instances for local agents (like the TPS) may require multiple instances.
Subsystems can be cloned, meaning they essentially are clustered, operating as a single unit, which is good for load balancing and high availability. Cloning is especially good for CAs and KRAs, where the same key and certificate data may need to be accessed by larger numbers of users or to have reliable failover.
When planning for multiple subsystem instances, keep in mind how the subsystems fit within the established security domains. Security domains create trusted relationships between subsystems, allowing them to work together to find available subsystems to respond to immediate needs. Multiple security domains can be used in a single PKI, with multiple instances of any kind of subsystem, or a single security domain can be used for all subsystems.
- Will any subsystems need to be cloned and, if so, what are the methods for securely storing their key materials?
- Cloned subsystems work together, essentially as a single instance. This can be good for high demand systems, failover, or load balancing, but it can become difficult to maintain. For example, cloned CAs have serial number ranges for the certificates they issue, and a clone could hit the end of its range.
- Will the subsystem certificates and keys be stored on the internal software token in Certificate System or on an external hardware token?
- Certificate System supports two hardware security modules (HSM): nCipher nShield and Safenet LunaSA. Using a hardware token can require additional setup and configuration before installing the subsystems, but it also adds another layer of security.
- What are the requirements for the CA signing certificate? Does the Certificate System need control over attributes like the validity period? How will the CA certificates be distributed?
The real question here is, will the CA bee a root CA which sets its own rules on issuing certificates or will it be a subordinate CA where another CA (another CA in your PKI or even an external CA) sets restrictions on what kind of certificates it can issue.
A Certificate Manager can be configured as either a root CA or a subordinate CA. The difference between a root CA and a subordinate CA is who signs the CA signing certificate. A root CA signs its own certificate. A subordinate CA has another CA (either internal or external) sign its certificate.
A self-signing root CA issues and signs its own CA signing certificate. This allows the CA to set its own configuration rules, like validity periods and the number of allowed subordinate CAs.
A subordinate CA has its certificates issued by a public CA or another Certificate System root CA. This CA is subordinate to the other CA’s rules about its certificate settings and how the certificate can be used, such as the kinds of certificates that it can issue, the extensions that it is allowed to include in certificates, and the levels of subordinate CAs the subordinate CA can create.
One option is to have the Certificate manager subordinate to a public CA. This can be very restrictive, since it introduces the restrictions that public CAs place on the kinds of certificates the subordinate CA can issue and the nature of the certificate chain. On the other hand, one benefit of chaining to a public CA is that the third party is responsible for submitting the root CA certificate to a web browser or other client software, which is a major advantage for certificates that are accessed by different companies with browsers that cannot be controlled by the administrator.
The other option is make the CA subordinate to a Certificate System CA. Setting up a Certificate System CA as the root CA means that the Certificate System administrator has control over all subordinate CAs by setting policies that control the contents of the CA signing certificates issued.
It is easiest to make the first CA installed a self-signed root, so that it is not necessary to apply to a third party and wait for the certificate to be issued. Make sure that you determine how many root CAs to have and where both root and subordinate CAs will be located.
- What kinds of certificates will be issued? What characteristics do they need to have, and what profile settings are available for those characteristics? What restrictions need to be placed on the certificates?
- As touched on in Section 5.4.6, “Using and customizing certificate profiles”, the profiles (the forms which configure each type of certificate issued by a CA) can be customized. This means that the subject DN, the validity period, the type of SSL/TLS client, and other elements can be defined by an administrator for a specific purpose. For security, profiles should provide only the functionality that is required by the PKI. Any unnecessary profiles should be disabled.
- What are the requirements for approving a certificate request? How does the requester authenticate himself, and what kind of process is required to approve the request?
The request approval process directly impacts the security of the certificate. Automated processes are much faster but are less secure since the identity of the requester is only superficially verified. Likewise, agent-approved enrollments are more secure (since, in the most secure scenario they can require an in-person verification and supporting identification) but they can also be the most time-consuming.
First determine how secure the identity verification process needs to be, then determine what authentication (approval) mechanism is required to validate that identity.
- Will many external clients need to validate certificate status? Can the internal OCSP in the Certificate Manager handle the load?
- Publishing CRLs and validating certificates is critical. Determine what kinds of clients will be checking the certificate status (will it mainly be internal clients? external clients? will they be validating user certificates or server certificates?) and also try to determine how many OCSP checks will be run. The CA has an internal OCSP which can be used for internal checks or infrequent checks, but a large number of OCSP checks could slow down the CA performance. For a larger number of OCSP checks and especially for large CRLs, consider using a separate OCSP Manager to take the load off the CA.
- Will the PKI allow replacement keys? Will it require key archival and recovery?
- If lost certificates or tokens are simply revoked and replaced, then there does not need to be a mechanism to recover them. However, when certificates are used to sign or encrypt data such as emails, files, or harddrives, then the key must always be available so that the data can be recovered. In that case, use a KRA to archive the keys so the data can always be accessed.
- Will the organization use smart cards? If so, will temporary smart cards be allowed if smart cards are mislaid, requiring key archival and recovery?
- If no smart cards are used, then it is never necessary to configure the TPS or TKS, since those subsystems are only used for token management. However, if smart cards are used, then the TPS and TKS are required. If tokens can be lost and replaced, then it is also necessary to have a KRA to archive the keys so that the token’s certificates can be regenerated.
- Where will certificates and CRLs be published? What configuration needs to be done on the receiving end for publishing to work? What kinds of certificates or CRLs need to be published and how frequently?
- The important thing to determine is what clients need to be able to access the certificate or CRL information and how that access is allowed. From there, you can define the publishing policy.
