How to Implement a Provider
in the
Java ™ Cryptography Architecture

Introduction

Who Should Read This Document

Related Documentation

Notes on Terminology

Engine Classes and Corresponding SPI Classes

Steps to Implement and Integrate a Provider

Step 1: Write your Service Implementation Code

Step 1.1: Additional JCA Provider Requirements and Recommendations for Encryption Implementations

Step 2: Give your Provider a Name

Step 3: Write your Master Class, a subclass of Provider

Step 3.1: Additional Steps for Cipher Implementations

Step 4: Compile your Code

Step 5: Place your Provider in a Jar File

Step 6: Optional -- Sign your JAR File

Step 6.1: Get a Code-Signing Certificate

Step 6.2: Sign Your Provider

Step 7: Prepare for Testing

Step 7.1: Install the Provider

Step 7.2: Set Provider Permissions

Step 8: Write and Compile your Test Programs

Step 9: Run your Test Programs

Step 10: Apply for U.S.Government Export Approval if Required

Step 11: Document your Provider and its Supported Services

Message Digests and MACs

Step 12: Make your Class Files and Documentation Available to Clients

How a Provider Can Do Self-Integrity Checking

Notes on the Sample Code

Finding the Provider JAR File: Basics

Determining the Provider's JAR File URL

Creating a JarFile Referring to the JAR File

Verifying the Provider JAR File: Basics

Verification Setup

JAR File Signature Check

Verifying Signatures

Ensuring Signers Are Trusted

Getting the List of Certificates

Identifying Each of the Signers and Determining If One is Trusted

Notes on myJCE Sample Code

Further Implementation Details and Requirements

Alias Names

Service Interdependencies

Default Initializations

Default Key Pair Generator Parameter Requirements

The Provider.Service Class

Signature Formats

DSA Interfaces and their Required Implementations

RSA Interfaces and their Required Implementations

Diffie-Hellman Interfaces and their Required Implementations

Interfaces for Other Algorithm Types

Algorithm Parameter Specification Interfaces and Classes

Key Specification Interfaces and Classes Required by Key Factories

Secret-Key Generation

Ensuring Exportability

Appendix A: The Sun Provider Master Class

Appendix B: The SunJCE Provider Master Class

Appendix C: The java.security Properties File

Introduction

The Java platform defines a set of APIs spanning major security areas, including cryptography, public key infrastructure, authentication, secure communication, and access control. These APIs allow developers to easily integrate security into their application code. They were designed around the following principles:

  1. Implementation independence

    Applications do not need to implement security themselves. Rather, they can request security services from the Java platform. Security services are implemented in providers (see below), which are plugged into the Java platform via a standard interface. An application may rely on multiple independent providers for security functionality.
  2. Implementation interoperability

    Providers are interoperable across applications. Specifically, an application is not bound to a specific provider, and a provider is not bound to a specific application.
  3. Algorithm extensibility

    The Java platform includes a number of built-in providers that implement a basic set of security services that are widely used today. However, some applications may rely on emerging standards not yet implemented, or on proprietary services. The Java platform supports the installation of custom providers that implement such services.

A Cryptographic Service Provider (provider) refers to a package (or a set of packages) that supply a concrete implementation of a subset of the cryptography aspects of the JDK Security API.

The java.security.Provider class encapsulates the notion of a security provider in the Java platform. It specifies the provider's name and lists the security services it implements. Multiple providers may be configured at the same time, and are listed in order of preference. When a security service is requested, the highest priority provider that implements that service is selected.

Figures 1 and 2 illustrate these options for requesting an MD5 message digest implementation. Both figures show three providers that implement message digest algorithms. The providers are ordered by preference from left to right (1-3). In Figure 1, an application requests an MD5 algorithm implementation without specifying a provider name. The providers are searched in preference order and the implementation from the first provider supplying that particular algorithm, ProviderB, is returned. In Figure 2, the application requests the MD5 algorithm implementation from a specific provider, ProviderC. This time the implementation from that provider is returned, even though a provider with a higher preference order, ProviderB, also supplies an MD5 implementation.

<Image of the General JCA Architecture>

Figure 1: Provider searching Figure 2: Specific provider requested

Figures 1 and 2 illustrate these options for requesting an MD5 message digest implementation. Both figures show three providers that implement message digest algorithms. The providers are ordered by preference from left to right (1-3). In Figure 1, an application requests an MD5 algorithm implementation without specifying a provider name. The providers are searched in preference order and the implementation from the first provider supplying that particular algorithm, ProviderB, is returned. In Figure 2, the application requests the MD5 algorithm implementation from a specific provider, ProviderC. This time the implementation from that provider is returned, even though a provider with a higher preference order, ProviderB, also supplies an MD5 implementation.

Each installation has one or more provider packages installed. Clients may configure their runtimes with different providers, and specify a preference order for each of them. The preference order is the order in which providers are searched for requested algorithms when no particular provider is requested.

Sun's version of the Java runtime environment comes standard with a default provider, named "SUN". Other Java runtime environments may not necessarily supply the "SUN" provider.

Who Should Read This Document

Programmers that only need to use the Java Security API to access existing cryptography algorithms and other services do not need to read this document.

This document is intended for experienced programmers wishing to create their own provider packages supplying cryptographic service implementations. It documents what you need to do in order to integrate your provider into Java so that your algorithms and other services can be found when Java Security API clients request them.

Related Documentation

This document assumes you have already read the Java Cryptography Architecture Reference Guide.

It documents the packages which contain the various classes and interfaces in the Security API.

Notes on Terminology

Throughout this document, the terms JCA by itself refers to the JCA framework. Whenever this document notes a specific JCA provider, it will be referred to explicitly by the provider name.

Engine Classes and Corresponding Service Provider Interface Classes

An engine class defines a cryptographic service in an abstract fashion (without a concrete implementation).

A cryptographic service is always associated with a particular algorithm or type. It either provides cryptographic operations (like those for digital signatures or message digests, ciphers or key agreement protocols ); generates or supplies the cryptographic material (keys or parameters) required for cryptographic operations; or generates data objects (keystores or certificates) that encapsulate cryptographic keys (which can be used in a cryptographic operation) in a secure fashion.

For example, here are four engine classes:

The Java Cryptography Architecture encompasses the classes comprising the Security package that relate to cryptography, including the engine classes. Users of the API request and utilize instances of the engine classes to carry out corresponding operations. The JDK defines the following engine classes:

Note:    A generator creates objects with brand-new contents, whereas a factory creates objects from existing material (for example, an encoding).

An engine class provides the interface to the functionality of a specific type of cryptographic service (independent of a particular cryptographic algorithm). It defines Application Programming Interface (API) methods that allow applications to access the specific type of cryptographic service it provides. The actual implementations (from one or more providers) are those for specific algorithms. For example, the Signature engine class provides access to the functionality of a digital signature algorithm. The actual implementation supplied in a SignatureSpi subclass (see next paragraph) would be that for a specific kind of signature algorithm, such as SHA1 with DSA, SHA1 with RSA, or MD5 with RSA.

The application interfaces supplied by an engine class are implemented in terms of a Service Provider Interface (SPI). That is, for each engine class, there is a corresponding abstract SPI class, which defines the Service Provider Interface methods that cryptographic service providers must implement.

An instance of an engine class, the "API object", encapsulates (as a private field) an instance of the corresponding SPI class, the "SPI object". All API methods of an API object are declared "final", and their implementations invoke the corresponding SPI methods of the encapsulated SPI object. An instance of an engine class (and of its corresponding SPI class) is created by a call to the getInstance factory method of the engine class.

The name of each SPI class is the same as that of the corresponding engine class, followed by "Spi". For example, the SPI class corresponding to the Signature engine class is the SignatureSpi class.

Each SPI class is abstract. To supply the implementation of a particular type of service and for a specific algorithm, a provider must subclass the corresponding SPI class and provide implementations for all the abstract methods.

Another example of an engine class is the MessageDigest class, which provides access to a message digest algorithm. Its implementations, in MessageDigestSpi subclasses, may be those of various message digest algorithms such as SHA-1, MD5, or MD2.

As a final example, the KeyFactory engine class supports the conversion from opaque keys to transparent key specifications, and vice versa. See Key Specification Interfaces and Classes Required by Key Factories for details. The actual implementation supplied in a KeyFactorySpi subclass would be that for a specific type of keys, e.g., DSA public and private keys.

Steps to Implement and Integrate a Provider

Follow the steps below to implement a provider and integrate it into the JCA framework:

Step 1: Write your Service Implementation Code

Step 1.1: Additional JCA Provider Requirements and Recommendations for Encryption Implementations

Step 2: Give your Provider a Name

Step 3: Write your Master Class, a subclass of Provider

Step 4: Compile your Code

Step 5: Place your Provider in a JAR file

Step 6: Optional -- Sign your JAR File

Step 6.1: Get a Code-Signing Certificate

Step 6.2: Sign your Provider

Step 7: Prepare for Testing

Step 7.1: Install the Provider

Step 7.2: Set Provider Permissions

Step 8: Write and Compile Test Programs

Step 9: Run your Test Programs

Step 10: Apply for U.S. Government Export Approval If Required

Step 11: Document your Provider and its Supported Services

Step 12: Make Your Class Files and Documentation Available to Clients

Step 1: Write your Service Implementation Code

The first thing you need to do is to write the code that provides algorithm-specific implementations of the cryptographic services you want to support.

Note that your provider may supply implementations of cryptographic services already available in one or more of the security components of the JDK.

For cryptographic services not defined in JCA (For example; signatures and message digests), please refer to Java Cryptography Architecture Reference Guide.

For each cryptographic service you wish to implement, create a subclass of the appropriate SPI class. JCA defines the following engine classes:

(See Engine Classes and Corresponding SPI Classes in this document for information on the JCA and other cryptographic classes.)

In your subclass, you need to:

  1. Supply implementations for the abstract methods, whose names usually begin with engine. See Further Implementation Details and Requirements for additional information.
  2. Ensure there is a public constructor without any arguments. Here's why: When one of your services is requested, Java Security looks up the subclass implementing that service, as specified by a property in your "master class" (see Step 3). Java Security then creates the Class object associated with your subclass, and creates an instance of your subclass by calling the newInstance method on that Class object. newInstance requires your subclass to have a public constructor without any parameters.
  3. A default constructor without arguments will automatically be generated if your subclass doesn't have any constructors. But if your subclass defines any constructors, you must explicitly define a public constructor without arguments.

Step 1a: Additional JCA Provider Requirements and Recommendations for Encryption Implementations

When instantiating a provider's implementation (class) of a Cipher, KeyAgreement, KeyGenerator, MAC or SecretKey factory, the framework will determine the provider's codebase (JAR file) and verify its signature. In this way, JCA authenticates the provider and ensures that only providers signed by a trusted entity can be plugged into JCA. Thus, one requirement for encryption providers is that they must be signed, as described in later steps.

In addition, each provider should perform self-integrity checking to ensure that the JAR file containing its code has not been manipulated in an attempt to invoke provider methods directly rather than through JCA. For further information, see How a Provider Can Do Self-Integrity Checking.

In order for provider classes to become unusable if instantiated by an application directly, bypassing JCA, providers should implement the following:

For providers that may be exported outside the U.S., CipherSpi implementations must include an implementation of the engineGetKeySize method which, given a Key, returns the key size. If there are restrictions on available cryptographic strength specified in jurisdiction policy files, each Cipher initialization method calls engineGetKeySize and then compares the result with the maximum allowable key size for the particular location and circumstances of the applet or application being run. If the key size is too large, the initialization method throws an exception.

Additional optional features that providers may implement are

Step 2: Give your Provider a Name

Decide on a name for your provider. This is the name to be used by client applications to refer to your provider.

Step 3: Write your Master Class, a subclass of Provider

The third step is to create a subclass of the java.security.Provider class.

Your subclass should be a final class, and its constructor should

In each of these, algName, certType, storeType, or attrName is the "standard" name of the algorithm, certificate type, keystore type, or attribute. (See Appendix A of the Java Cryptography Architecture Reference Guide for the standard names that should be used.)

For a property in the above format, the value of the property must be the value for the corresponding attribute. (See Appendix A of the Java Cryptography Architecture API Specification & Reference for the definition of each standard attribute.)

As an example, the default provider named "SUN" implements the SHA1withDSA Digital Signature Algorithm in software. In the master class for the provider "SUN", it sets the Signature.SHA1withDSA ImplementedIn to have the value Software via the following:

    put("Signature.SHA1withDSA ImplementedIn", "Software")

For further master class property setting examples, see Appendix A to view the current Sun.java source file or Appendix B to see the SunJCE provider. These files show how the Sun and SunJCE providers set properties.

Step 3.1: Additional Steps for Cipher Implementations

As mentioned above, in the case of a Cipher property, algName may actually represent a transformation. A transformation is a string that describes the operation (or set of operations) to be performed by a Cipher object on some given input. A transformation always includes the name of a cryptographic algorithm (e.g., DES), and may be followed by a mode and a padding scheme.

A transformation is of the form:

(In the latter case, provider-specific default values for the mode and padding scheme are used). For example, the following is a valid transformation:

    Cipher c = Cipher.getInstance("DES/CBC/PKCS5Padding");

When requesting a block cipher in stream cipher mode (for example; DES in CFB or OFB mode), a client may optionally specify the number of bits to be processed at a time, by appending this number to the mode name as shown in the following sample transformations:

    Cipher c1 = Cipher.getInstance("DES/CFB8/NoPadding");
    Cipher c2 = Cipher.getInstance("DES/OFB32/PKCS5Padding");

If a number does not follow a stream cipher mode, a provider-specific default is used. (For example, the SunJCE provider uses a default of 64 bits.)

A provider may supply a separate class for each combination of algorithm/mode/padding. Alternatively, a provider may decide to provide more generic classes representing sub-transformations corresponding to algorithm or algorithm/mode or algorithm//padding (note the double slashes); in this case the requested mode and/or padding are set automatically by the getInstance methods of Cipher, which invoke the engineSetMode and engineSetPadding methods of the provider's subclass of CipherSpi.

That is, a Cipher property in a provider master class may have one of the formats shown in the table below.

Cipher Property Format Description
Cipher.algName A provider's subclass of CipherSpi implements algName with pluggable mode and padding
Cipher.algName/mode A provider's subclass of CipherSpi implements algName in the specified mode, with pluggable padding
Cipher.algName//padding A provider's subclass of CipherSpi implements algName with the specified padding, with pluggable mode
Cipher.algName/mode/padding A provider's subclass of CipherSpi implements algName with the specified mode and padding

(See Appendix A of the Java Cryptography Architecture Reference Guide for the standard algorithm names, modes, and padding schemes that should be used.)

For example, a provider may supply a subclass of CipherSpi that implements DES/ECB/PKCS5Padding, one that implements DES/CBC/PKCS5Padding, one that implements DES/CFB/PKCS5Padding, and yet another one that implements DES/OFB/PKCS5Padding. That provider would have the following Cipher properties in its master class:

Another provider may implement a class for each of the above modes (i.e., one class for ECB, one for CBC, one for CFB, and one for OFB), one class for PKCS5Padding, and a generic DES class that subclasses from CipherSpi. That provider would have the following Cipher properties in its master class:

The getInstance factory method of the Cipher engine class follows these rules in order to instantiate a provider's implementation of CipherSpi for a transformation of the form "algorithm":

  1. Check if the provider has registered a subclass of CipherSpi for the specified "algorithm".

    If the answer is YES, instantiate this class, for whose mode and padding scheme default values (as supplied by the provider) are used.

    If the answer is NO, throw a NoSuchAlgorithmException exception.

The getInstance factory method of the Cipher engine class follows these rules in order to instantiate a provider's implementation of CipherSpi for a transformation of the form "algorithm/mode/padding":

  1. Check if the provider has registered a subclass of CipherSpi for the specified "algorithm/mode/padding" transformation.

    If the answer is YES, instantiate it.

    If the answer is NO, go to the next step.

  2. Check if the provider has registered a subclass of CipherSpi for the sub-transformation "algorithm/mode".

    If the answer is YES, instantiate it, and call engineSetPadding(padding) on the new instance.

    If the answer is NO, go to the next step.
  3. Check if the provider has registered a subclass of CipherSpi for the sub-transformation "algorithm//padding" (note the double slashes).

    If the answer is YES, instantiate it, and call engineSetMode(mode) on the new instance.

    If the answer is NO, go to the next step.
  4. Check if the provider has registered a subclass of CipherSpi for the sub-transformation "algorithm".

    If the answer is YES, instantiate it, and call engineSetMode(mode) and engineSetPadding(padding) on the new instance.

    If the answer is NO, throw a NoSuchAlgorithmException exception.

Step 4: Compile your Code

After you have created your implementation code (Step 1), given your provider a name (Step 2), and created the master class (Step 3), use the Java compiler to compile your files.

Step 5: Place Your Provider in a JAR File

Place your provider code in a JAR file, in preparation for signing it in the next step. For more information on the jar tool, see jar (for Solaris) (for Microsoft Windows). 

    jar cvf <JAR file name> <list of classes, separated by spaces>

This command creates a JAR file with the specified name containing the specified classes.

Step 6: Optional -- Sign your JAR File

If your provider is supplying encryption algorithms through the Cipher KeyAgreement, KeyGenerator, Mac, or SecretKeyFactory classes, you will need to sign your JAR file so that the JCA can authenticate the code at runtime. For details, see Step 1a. If you are NOT providing an implementation of this type you can skip this step.

Step 6.1: Get a Code-Signing Certificate

The next step is to request a code-signing certificate so that you can use it to sign your provider prior to testing. The certificate will be good for both testing and production. It will be valid for 5 years.

Below are the steps you should use to get a code-signing certificate. For more information on the keytool tool, see keytool (for Solaris) (for Microsoft Windows).

  1. Use keytool to generate a DSA keypair, using DSA algorithm as an example: 
        keytool -genkeypair -alias <alias> \
	-keyalg DSA -keysize 1024 \
	-dname "cn=<Company Name>, \
	ou=Java Software Code Signing,\
	o=Sun Microsystems Inc" \
	-keystore <keystore file name>\
	-storepass <keystore password>
    This will generate a DSA keypair (a public key and an associated private key) and store it in an entry in the specified keystore. The public key is stored in a self-signed certificate. The keystore entry can subsequently be accessed using the specified alias.

    The option values in angle brackets ("<" and ">") represent the actual values that must be supplied. For example, <alias> must be replaced with whatever alias name you wish to be used to refer to the newly-generated keystore entry in the future, and <keystore file name> must be replaced with the name of the keystore to be used. Note: Do not surround actual values with angle brackets. For example, if you want your alias to be myTestAlias, specify the -alias option as follows: 

        -alias myTestAlias
    If you specify a keystore that doesn't yet exist, it will be created.

    Note: If command lines you type are not allowed to be as long as the keytool -genkeypair command you want to execute (for example, if you are typing to a Microsoft Windows DOS prompt), you can create and execute a plain-text batch file containing the command. That is, create a new text file that contains nothing but the full keytool -genkeypair command. (Remember to type it all on one line.) Save the file with a .bat extension. Then in your DOS window, type the file name (with its path, if necessary). This will cause the command in the batch file to be executed.

  2. Use keytool to generate a certificate signing request. 
        keytool -certreq -alias <alias> \
	-file <csr file name> \
	-keystore <keystore file name> \
	-storepass <keystore password>
    Here, <alias> is the alias for the DSA keypair entry created in the previous step. This command generates a Certificate Signing Request (CSR), using the PKCS#10 format. It stores the CSR in the file whose name is specified in <csr file name>.
  3. Send the CSR, contact information, and other required documentation to the JCA Code Signing Certification Authority. Send, via email, the CSR and contact information (see below) to javasoft-cert-request@sun.com. Put the following in the Subject line of your email message: 
        Request a Certificate for Signing a JCA Provider
    Put the contact information in the body of the message and send the CSR file as a plain text attachment to the message. If your mail tool has an option for specifying the encoding format to be used for attachments, select the "MIME" option.

    Note: The CSR file is just a plain text file, in Base 64 encoding. Only the first and last lines are human-readable.

    Include the following contact information in the body of your message:
    • Company Name
    • Street Address (Not a post office box)
    • City
    • State/Province
    • Country
    • Company Telephone Number
    • Company Fax Number
    • Requester Name
    • Requester Telephone Number
    • Requester Email Address
    • Brief description of your company (size, line of business, etc.)

    All of the above information is required.

  4. After the JCA Code Signing Certification Authority has received your email message, they will send you a request number via email. Please allow five business days for processing.

    Once you receive this request number, you should print, fill out and mail the Certification Form for CSPs. This form should be mailed to the address below. Be sure to include the request number on the form so that your hardcopy mailing can be matched to the email message containing your CSR and contact information.

    Sun Microsystems, Inc.
    International Trade Services/Export Compliance
    Attn: Encryption Export
    10 Network Circle MS: UMPK10-144 Menlo Park, CA 94025 .

    You can also fax paperwork as follows:

    Fax #: 408-668-0918

    Subject Line: CSR #___
    Attention: Encryption Export
    Sun Microsystems, Inc.
    International Trade Services/Export Compliance

    After the JCA Code Signing Certification Authority has received both your email message and the required form, they will authenticate you, the requester. Then they will create and sign a code-signing certificate valid for 5 years. You will receive an email message containing two plain-text file attachments: one file containing this code-signing certificate and another file containing its own CA certificate, which authenticates its public key. Please also allow five business days from receipt of your request for processing.

  5. Use keytool to import the certificates received from the CA.

    Once you have received the two certificates from the JCA Code Signing Certification Authority, you can use keytool to import them into your keystore.

    First import the CA's certificate as a "trusted certificate": 
        keytool -import -alias <alias for the CA cert> \
	-file <CA cert file name> \
	-keystore <keystore file name> \
	-storepass <keystore password>
    Then import the code-signing certificate: 
        keytool -import -alias <alias> \
	-file <code-signing cert file name> \
	-keystore <keystore file name> \
	-storepass <keystore password>
    Here, <alias> is the same alias as that which you created in step 1 where you generated a DSA keypair. This command replaces the self-signed certificate in the keystore entry specified by <alias> with the one signed by the JCA Code Signing Certification Authority.

Now that you have in your keystore a certificate from an entity trusted by JCA (the JCA Code Signing Certification Authority), you can place your provider code in a JAR file (Step 5) and then use that certificate to sign the JAR file (Step 6.2).

Step 6.2: Sign Your Provider

Sign the JAR file created step five with the code-signing certificate obtained in Step 6. For more information on the jarsigner tool, see jarsigner (for Solaris) (for Microsoft Windows). 

    jarsigner -keystore <keystore file name> \
	-storepass <keystore password> \
	<JAR file name> <alias>

Here, <alias> is the alias into the keystore for the entry containing the code-signing certificate received from the JCA Code Signing Certification Authority (the same alias as that specified in the commands in Step 6.1).

You can test verification of the signature via the following: 

    jarsigner -verify <JAR file name>

The text "jar verified" will be displayed if the verification was successful.

Step 7: Prepare for Testing

The next steps describe how to install and configure your new provider so that it is available via the JCA.

Step 7.1: Install the Provider

In order to prepare for testing your provider, you must install it in the same manner as will be done by clients wishing to use it. The installation enables Java Security to find your algorithm implementations when clients request them.

Installing a provider is done in two steps: installing the provider package classes, and configuring the provider.

Installing the Provider Classes

The first thing you must do is make your classes available so that they can be found when requested. You ship your provider classes as a JAR (Java ARchive) file.

There are a two possible ways to install provider classes:

The provider JAR file will be considered an installed extension if it is placed in the standard place for the JAR files of an installed extension:

    <java-home>/lib/ext		[Solaris]
    <java-home>\lib\ext		[Windows]

Here <java-home> refers to the directory where the runtime software is installed, which is the top-level directory of the JavaTM 2 Runtime Environment (JRE) or the jre directory in the JavaTM SE (JDK) software. For example, if you have JDK 6 installed on Solaris in a directory named /home/user1/jdk1.6.0, or on Microsoft Windows in a directory named C:\jdk1.6.0, then you need to install the JAR file in the following directory: 

    /home/user1/jdk1.6.0/jre/lib/ext	[Solaris]
    C:\jdk1.6.0\jre\lib\ext		[Windows]

Similarly, if you have JRE 6 installed on Solaris in a directory named /home/user1/jre1.6.0, or on Microsoft Windows in a directory named C:\jre1.6.0, you need to install the JAR file in the following directory: 

    /home/user1/jre1.6.0/lib/ext	[Solaris]
    C:\jre1.6.0\lib\ext			[Windows]

For more information on installed extensions, see Installed Extensions.

For more information on bundled extensions, see Bundled Extensions.

Configuring the Provider

The next step is to add the provider to your list of approved providers. This is done statically by editing the security properties file 

    <java-home>/lib/security/java.security	[Solaris]
    <java-home>\lib\security\java.security	[Windows]

Here <java-home> refers to the directory where the JRE was installed. For example, if you have JDK 6 installed on Solaris in a directory named /home/user1/jdk1.6.0, or on Microsoft indows in a directory named C:\jdk1.6.0, then you need to edit the following file: 

    /home/user1/jdk1.6.0/jre/lib/security/java.security	[Solaris]
    C:\jdk1.6.0\jre\lib\security\java.security		[Windows]

Similarly, if you have the JRE 6 installed on Solaris in a directory named /home/user1/jre1.6.0, or on Windows in a directory named C:\jre1.6.0, then you need to edit this file: 

    /home/user1/jre1.6.0/lib/security/java.security	[Solaris]
    C:\jre1.6.0\lib\security\java.security		[Windows]

For each provider, this file should have a statement of the following form: 

    security.provider.n=masterClassName

This declares a provider, and specifies its preference order n. The preference order is the order in which providers are searched for requested algorithms when no specific provider is requested. The order is 1-based; 1 is the most preferred, followed by 2, and so on.

masterClassName must specify the fully qualified name of the provider's "master class", which you implemented in Step 3. This class is always a subclass of the Provider class.

Java comes standard with providers named SUN, SunRsaSign, and SunJCE which are automatically configured as a static provider in the java.security properties file, as follows: 

    security.provider.2=sun.security.provider.Sun
    security.provider.3=sun.security.rsa.SunRsaSign
    security.provider.4=sun.security.provider.SunJCE

(The Sun provider's master class is the Sun class in the sun.security.provider package.)

The JCA provider SunJCE and other security-related providers shipped with the Java platform are also automatically configured as static providers.

To utilize another JCA provider, add a line registering the alternate provider, giving it a lower preference order than the SUN and SunRsaSign providers.

Suppose that your master class is the CryptoX class in the com.cryptox.provider package, and that you would like to make your provider the fourth preferred provider. To do so, edit the java.security file as seen below:

    security.provider.2=sun.security.provider.Sun
    security.provider.3=sun.security.rsa.SunRsaSign
    security.provider.4=com.cryptox.provider.CryptoX
    security.provider.5=sun.security.provider.SunJCE

Note: Providers may also be registered dynamically. To do so, a program (such as your test program, to be written in Step 8) can call either the addProvider or insertProviderAt method in the Security class. This type of registration is not persistent and can only be done by code which is granted the following permission:

    java.security.SecurityPermission "insertProvider.{name}"

where {name} is replaced by the actual provider name.

For example, if the provider name is "MyJCE" and if the provider's code is in the myjce_provider.jar file in the /localWork directory, then here is a sample policy file grant statement granting that permission: 

    grant codeBase "file:/localWork/myjce_provider.jar" {
	permission java.security.SecurityPermission
	    "insertProvider.MyJCE";
    };

Step 7.2: Set Provider Permissions

Whenever providers are not installed extensions, permissions must be granted for when applets or applications are run while a security manager is installed. There is typically a security manager installed whenever an applet is running, and a security manager may be installed for an application either via code in the application itself or via a command-line argument. Permissions do not need to be granted to installed extensions, since the default system policy file grants all permissions to installed extensions.

Whenever a client does not install your provider as an installed extension, your provider may need the following permissions granted to it in the client environment:

To ensure your provider works when a security manager is installed and the provider is not an installed extension, you need to test such an installation and execution environment. In addition, prior to testing you need to grant appropriate permissions to your provider and to any other providers it uses. For example, a sample statement granting permissions to a provider whose name is "MyJCE" and whose code is in myjce_provider.jar appears below. Such a statement could appear in a policy file. In this example, the myjce_provider.jar file is assumed to be in the /localWork directory. 

    grant codeBase "file:/localWork/myjce_provider.jar" {
	permission java.lang.RuntimePermission "getProtectionDomain";
	permission java.security.SecurityPermission
	    "putProviderProperty.MyJCE";
    };

Step 8: Write and Compile your Test Programs

Write and compile one or more test programs that test your provider's incorporation into the Security API as well as the correctness of its algorithm(s). Create any supporting files needed, such as those for test data to be encrypted.

The first tests your program should perform are ones to ensure that your provider is found, and that its name, version number, and additional information is as expected. To do so, you could write code like the following, substituting your provider name for MyPro: 

    import java.security.*;

    Provider p = Security.getProvider("MyPro");

    System.out.println("MyPro provider name is " + p.getName());
    System.out.println("MyPro provider version # is " + p.getVersion());
    System.out.println("MyPro provider info is " + p.getInfo());

Next, you should ensure that your services are found. For instance, if you implemented the DES encryption algorithm, you could check to ensure it's found when requested by using the following code (again substituting your provider name for "MyPro"): 

    Cipher c = Cipher.getInstance("DES", "MyPro");

    System.out.println("My Cipher algorithm name is " + c.getAlgorithm());

If you don't specify a provider name in the call to getInstance, all registered providers will be searched, in preference order (see Configuring the Provider), until one implementing the algorithm is found.

If your provider implements an exemption mechanism, you should write a test applet or application that uses the exemption mechanism. Such an applet/application also needs to be signed, and needs to have a "permission policy file" bundled with it. See How to Make Applications "Exempt" from Cryptographic Restrictions in the Java Cryptography Architecture Reference Guide for complete information on creating and testing such an application.

Step 9: Run your Test Programs

Run your test program(s). Debug your code and continue testing as needed. If the Java Security API cannot seem to find one of your algorithms, review the steps above and ensure they are all completed.

Be sure to include testing of your programs using different installation options (e.g. making the provider an installed extension or placing it on the class path) and execution environments (with or without a security manager running). Installation options are discussed in Step 7.1. In particular, you need to ensure your provider works when a security manager is installed and the provider is not an installed extension -- and thus the provider must have permissions granted to it; therefore, you need to test such an installation and execution environment, after granting required permissions to your provider and to any other providers it uses, as described in Step 7.2.

If you find during testing that your code needs modification, make the changes, recompile (Step 4), place the updated provider code in a JAR file (Step 6), sign the JAR file if necessary (Step 6.2), re-install the provider (Step 7.1), if needed fix or add to the permissions (Step 7.2), and then re-test your programs. Repeat these steps as needed.

Step 10: Apply for U.S. Government Export Approval If Required

All U.S. vendors whose providers may be exported outside the U.S. should apply to the Bureau of Industry and Security in the U.S. Department of Commerce for export approval. Please consult your export counsel for more information.

Note: If your provider calls Cipher.getInstance() and the returned Cipher object needs to perform strong cryptography regardless of what cryptographic strength is allowed by the user's downloaded jurisdiction policy files, you should include a copy of the cryptoPerms permission policy file which you intend to bundle in the JAR file for your provider and which specifies an appropriate permission for the required cryptographic strength. The necessity for this file is just like the requirement that applets and applications "exempt" from cryptographic restrictions must include a cryptoPerms permission policy file in their JAR file. For more information on the creation and inclusion of such a file, see How to Make Applications "Exempt" from Cryptographic Restrictions in the Java Cryptography Architecture Reference Guide.

Here are two URLs that may be useful:

Step 11: Document your Provider and its Supported Services

The next step is to write documentation for your clients. At the minimum, you need to specify:

In addition, your documentation should specify anything else of interest to clients, such as any default algorithm parameters.

Message Digests and MACs

For each Message Digest and MAC algorithm, indicate whether or not your implementation is cloneable. This is not technically necessary, but it may save clients some time and coding by telling them whether or not intermediate Message Digests or MACs may be possible through cloning. Clients who do not know whether or not a MessageDigest or Mac implementation is cloneable can find out by attempting to clone the object and catching the potential exception, as illustrated by the following example: 

    try {
	// try and clone it
	/* compute the MAC for i1 */
	mac.update(i1);
	byte[] i1Mac = mac.clone().doFinal();

	/* compute the MAC for i1 and i2 */
	mac.update(i2);
	byte[] i12Mac = mac.clone().doFinal();

	/* compute the MAC for i1, i2 and i3 */
	mac.update(i3);
	byte[] i123Mac = mac.doFinal();
    } catch (CloneNotSupportedException cnse) {
	// have to use an approach not involving cloning
    }

where:

  • mac is the MAC object they received when they requested one via a call to Mac.getInstance,

  • i1, i2 and i3 are input byte arrays, and

  • they want to calculate separate hashes for:
    • i1
    • i1 and i2
    • i1, i2, and i3

Key Pair Generators

For a key pair generator algorithm, in case the client does not explicitly initialize the key pair generator (via a call to an initialize method), each provider must supply and document a default initialization. For example, the Diffie-Hellman key pair generator supplied by the SunJCE provider uses a default prime modulus size (keysize) of 1024 bits.

Key Factories

A provider should document all the key specifications supported by its (secret-)key factory.

Algorithm Parameter Generators

In case the client does not explicitly initialize the algorithm parameter generator (via a call to an init method in the AlgorithmParameterGenerator engine class), each provider must supply and document a default initialization. For example, the SunJCE provider uses a default prime modulus size (keysize) of 1024 bits for the generation of Diffie-Hellman parameters, the Sun provider a default modulus prime size of 1024 bits for the generation of DSA parameters.

Signature Algorithms

If you implement a signature algorithm, you should document the format in which the signature (generated by one of the sign methods) is encoded. For example, the SHA1withDSA signature algorithm supplied by the "SUN" provider encodes the signature as a standard ASN.1 SEQUENCE of two integers, r and s.

Random Number Generation (SecureRandom) Algorithms

For a random number generation algorithm, provide information regarding how "random" the numbers generated are, and the quality of the seed when the random number generator is self-seeding. Also note what happens when a SecureRandom object (and its encapsulated SecureRandomSpi implementation object) is deserialized: If subsequent calls to the nextBytes method (which invokes the engineNextBytes method of the encapsulated SecureRandomSpi object) of the restored object yield the exact same (random) bytes as the original object would, then let users know that if this behaviour is undesirable, they should seed the restored random object by calling its setSeed method.

Certificate Factories

A provider should document what types of certificates (and their version numbers, if relevant), can be created by the factory.

Keystores

A provider should document any relevant information regarding the keystore implementation, such as its underlying data format.

Step 12: Make your Class Files and Documentation Available to Clients

After writing, configuring, testing, installing and documenting your provider software, make documentation available to your customers.

How a Provider Can Do Self-Integrity Checking

Each provider should do self-integrity checking to ensure that the JAR file containing its code has not been tampered with, for example in an attempt to invoke provider methods directly rather than through JCA. Providers that provide implementations for encryption services (Cipher, KeyAgreement, KeyGenerator, MAC or SecretKey factory) must be digitally signed and should be signed with a certificate issued by "trusted" Certification Authorities. Currently, the following two Certification Authorities are considered "trusted":

Please refer to Step 6.2 for detailed information on how to get a code-signing certificate from Sun Microsystems' JCA Code Signing CA and the certificate of that CA.

After getting the signing certificate from above Certification Authority, provider packages should embed within themselves the bytes for its own signing certificate, for example in an array like the bytesOfProviderCert array referred to in the Identifying Each of the Signers and Determining If One is Trusted section below. At runtime, the embedded certificate will be used in determining whether or not the provider code is authentic.

The basic approach a provider can use to check its own integrity is:

  1. Determine the URL of the JAR file containing the provider code, and
  2. Verify the JAR file's digital signatures to ensure that at least one signer of each entry of the JAR file is trusted.

Each of these steps is described in the following sections:

Notes on the Sample Code

Finding the Provider JAR File: Basics

Determining the Provider's JAR File URL

Creating a JarFile Referring to the JAR File

Verifying the Provider JAR File: Basics

Verification Setup

JAR File Signature Check

Verifying Signatures

Ensuring Signers Are Trusted

Getting the List of Certificates

Identifying Each of the Signers and Determining If One is Trusted

Notes on myJCE Sample Code



Notes on the Sample Code

Note: The sample code MyJCE.java is a complete code example that implements these steps. You can download this code for your reference. The Notes on the Sample Code section traces how these concepts are implemented in the sample code.

IMPORTANT NOTE: In the unbundled version of JCE 1.2.x, (used with JDKs 1.2.x and 1.3.x), providers needed to include code to authenticate the JCA framework to assure themselves of the integrity and authenticity of the JCA that they plugged into. In JDK 6, this is no longer necessary.

One implication is that a provider written just for JCE 1.2.2 will not work in JDK 6 because the provider's JCE framework authentication check will not work; the JCE framework code is no longer where the provider expects it to be. If you want your provider to work only with the JDK 6, it should not have code to authenticate the JCE framework. On the other hand, if you want your provider to work both with JCE 1.2.2 and with the JDK 6, then add a conditional statement. This way the provider code to authenticate the JCE framework is executed only when the provider is run with JCE 1.2.2. The following is sample code: 

    Class cipherCls = Class.forName("javax.crypto.Cipher");

    CodeSource cs = cipherCls.getProtectionDomain().getCodeSource();
    if (cs != null) {
	// Authenticate JCE framework

	 . . .
    }

Finding the Provider JAR File: Basics

Determining the Provider's JAR File URL

The URL for the provider's JAR file can be obtained by determining the provider's CodeSource and then calling the getLocation method on the CodeSource. 

    URL providerURL = (URL) AccessController.doPrivileged(
	new PrivilegedAction) {
	    public Object run() {
		CodeSource cs =
		    MyJCE.class.getProtectionDomain().getCodeSource();
		return cs.getLocation();
	    }
	});

Creating a JarFile Referring to the JAR File

Once you have the URL for the provider's JAR file, you can create a java.util.jar.JarFile referring to the JAR file. This instance is needed in the step for verifying the Provider JAR file.

To create the JAR file, first open a connection to the specified URL by calling its openConnection method. Since the URL is a JAR URL, the type is java.net.JarURLConnection. Here's the basic code: 

    // Prep the url with the appropriate protocol.
    jarURL =
	url.getProtocol().equalsIgnoreCase("jar") ? url :
	    new URL("jar:" + url.toString() + "!/");

    // Retrieve the jar file using JarURLConnection
    JarFile jf = (JarFile) AccessController.doPrivileged(
	new PrivilegedExceptionAction() {
	    public Object run() throws Exception {
		JarURLConnection conn =
		    (JarURLConnection) jarURL.openConnection();
	...

Now that you have a JarURLConnection, you can call its getJarFile method to get the JAR file:

    // Always get a fresh copy, so we don't have to
    // worry about the stale file handle when the
    // cached jar is closed by some other application.
    conn.setUseCaches(false);
    jf = conn.getJarFile();

Verifying the Provider JAR File: Basics

Once you have determined the URL for your provider JAR file and you have created a JarFile referring to the JAR file, as shown in the steps above, you can then verify the file.

The basic approach is:

  1. Ensure that at least one of each entry's signer's certificates is equal to the provider's own code signing certificate.
  2. Go through all the entries in the JAR file and ensure the signature on each one verifies correctly.
  3. Ensure that at least one of each entry's signer's certificates can be traced back to a trusted Certification Authority.

Sample code for each of these steps is presented and described in the following sections:

Verification Setup

JAR File Signature Check

Verifying Signatures

Ensuring Signers Are Trusted

Getting the List of Certificates

Identifying Each of the Signers and Determining If One is Trusted


Verification Setup

Our approach is to define a class JarVerifier to handle the retrieval of a JAR file from a given URL and verify whether the JAR file is signed with the specified certificate.

The constructor of JarVerifier takes the provider URL as a parameter which will be used to retrieve the JAR file later.

The actual jar verification is implemented in the verify method which takes the provider code signing certificate as a parameter. 

    public void verify(X509Certificate targetCert) throws IOException {
	// variable 'jarFile' is a JarFile object created
	// from the provider's Jar URL.
	...
	Vector entriesVec = new Vector();

Basically the verify method will go through the JAR file entries twice: the first time checking the signature on each entry and the second time verifying the signer is trusted.

Note: In our code snippets the jarFile variable is the JarFile object of the provider's jar file.

JAR File Signature Check

An authentic provider JAR file is signed. So the JAR file has been tampered with if it isn't signed: 

    // Ensure the jar file is signed.
    Manifest man = jarFile.getManifest();
    if (man == null) {
	throw new SecurityException("The provider is not signed");
    }

Verifying Signatures

The next step is to go through all the entries in the JAR file and ensure the signature on each one verifies correctly. One possible way to verify the signature on a JAR file entry is to simply read the file. If a JAR file is signed, the read method itself automatically performs the signature verification. Here is sample code: 

    // Ensure all the entries' signatures verify correctly
    byte[] buffer = new byte[8192];
    Enumeration entries = jarFile.entries();

    while (entries.hasMoreElements()) {
	JarEntry je = (JarEntry) entries.nextElement();

	// Skip directories.
	if (je.isDirectory())
	    continue;

	entriesVec.addElement(je);
	InputStream is = jarFile.getInputStream(je);

	// Read in each jar entry. A security exception will
	// be thrown if a signature/digest check fails.
	int n;
	while ((n = is.read(buffer, 0, buffer.length)) != -1) {
	    // Don't care
	}
	is.close();
    }

Ensuring Signers Are Trusted

The code in the previous section verified the signatures of all the provider JAR file entries. The fact that they all verify correctly is a requirement, but it is not sufficient to verify the authenticity of the JAR file. A final requirement is that the signatures were generated by the same entity as the one that developed this provider. To test that the signatures are trusted, we can again go through each entry in the JAR file (this time using the entriesVec built in the previous step), and for each entry that must be signed (that is, each entry that is not a directory and that is not in the META-INF directory):

  1. Get the list of signer certificates for the entry.
  2. Identify each of the certificate chains and determine whether any of the certificate chains are trusted. At least one of the certificate chains must be trusted.

The loop setup is the following:

    Enumeration e = entriesVec.elements();
    while (e.hasMoreElements()) {
	JarEntry je = (JarEntry) e.nextElement();
	...
    }

Getting the List of Certificates

The certificates for the signers of a JAR file entry JarEntry can be obtained simply by calling the JarEntry getCertificates method: 

    Certificate[] certs = je.getCertificates();

Adding this line of code to the previous loop setup code, and adding code to ignore directories and files in the META-INF directory gives us: 

    while (e.hasMoreElements()) {
	JarEntry je = (JarEntry) e.nextElement();

	// Every file must be signed except files in META-INF.
	Certificate[] certs = je.getCertificates();
	if ((certs == null) || (certs.length == 0)) {
	    if (!je.getName().startsWith("META-INF"))
		throw new SecurityException(
		    "The provider has unsigned class files.");
	    } else {
		// Check whether the file is signed by the expected
		// signer. The jar may be signed by multiple signers.
		// See if one of the signers is 'targetCert'.
		...
	    }
	...

Identifying Each of the Signers and Determining If One is Trusted

The certificate array returned by the JarEntry getCertificates method contains one or more certificate chains. There is one chain per signer of the entry. Each chain contains one or more certificates. Each certificate in a chain authenticates the public key in the previous certificate.

The first certificate in a chain is the signer's certificate which contains the public key corresponding to the private key actually used to sign the entry. Each subsequent certificate is a certificate for the issuer of the previous certificate. Since the self-integrity check is based on whether the JAR file is signed with the provider's signing cert, the trust decision will be made upon only the first certificate, the signer's certificate.

We need to go through the array of certificate chains and check each chain and the associated signers until we find a trusted entity. For each JAR file entry, at least one of the signers must be trusted. A signer is considered "trusted" if and only if its certificate is equals to the embedded provider signing certificate.

The following sample code loops through all the certificate chains, compares the first certificate in a chain to the embedded provider signing certificate, and only returns true if a match is found. 

    int startIndex = 0;
    X509Certificate[] certChain;
    boolean signedAsExpected = false;

    while ((certChain = getAChain(certs, startIndex)) != null) {
	if (certChain[0].equals(targetCert)) {
	    // Stop since one trusted signer is found.
	    signedAsExpected = true;
	    break;
	}

	// Proceed to the next chain.
	startIndex += certChain.length;
    }

    if (!signedAsExpected) {
	throw new SecurityException(
	    "The provider is not signed by a trusted signer");
    }

The getAChain method is defined as follows:

    /**
     * Extracts ONE certificate chain from the specified certificate array
     * which may contain multiple certificate chains, starting from index
     * 'startIndex'.
     */
    private static X509Certificate[] getAChain(
	    Certificate[] certs, int startIndex) {

	if (startIndex > certs.length - 1)
	    return null;

	int i;
	// Keep going until the next certificate is not the
	// issuer of this certificate.
	for (i = startIndex; i < certs.length - 1; i++) {
	    if (!((X509Certificate)certs[i + 1]).getSubjectDN().
		    equals(((X509Certificate)certs[i]).getIssuerDN())) {
		break;
	    }
	}

	// Construct and return the found certificate chain.
	int certChainSize = (i-startIndex) + 1;
	X509Certificate[] ret = new X509Certificate[certChainSize];
	for (int j = 0; j < certChainSize; j++ ) {
	    ret[j] = (X509Certificate) certs[startIndex + j];
	}
	return ret;
    }

Notes on the myJCE Code Sample

The sample code, MyJCE.java, is a sample provider which has a method selfIntegrityChecking which performs self-integrity checking. It first determines the URL of its own provider JAR file and then verifies that the provider JAR file is signed with the embedded code-signing certificate.

Note: The method selfIntegrityChecking should be called by all the constructors of its cryptographic engine classes to ensure that its integrity is not compromised.

Provider MyJCE performs self-integrity checking in the following steps:

  1. Determine the URL to access the provider JAR file using its own class, MyJCE.class.

  2. Instantiate a JarVerifier object with the provider URL in Step 1.

  3. Create a X509Certificate object from the embedded byte array bytesOfProviderCert.

  4. Call the JarVerifier.verify method to verify all entries in the provider JAR file are signed and are signed with the same certificate instantiated in Step 3.

Note: The class JarVerifier will retrieve the JAR file from the given URL, make sure the JAR file is signed, all entries have valid signatures, and that entries are signed with the specified X509Certificate.

A security exception is thrown by JarVerifier.verify in several cases:

The MyJCE.java sample code is comprised of the code snippets shown above. In addition, it includes error handling, sample code signing certificate bytes, and code for instantiating a X509Certificate object from the embedded sample code signing certificate bytes.

Regarding the use of AccessController.doPrivileged, please see API For Privileged Blocks for information on the use of doPrivileged.

Further Implementation Details and Requirements

Alias Names

For many cryptographic algorithms and types, there is a single official "standard name" defined in Appendix A of the Java Cryptography Architecture Reference Guide.

For example, "MD5" is the standard name for the RSA-MD5 Message Digest algorithm defined by RSA DSI in RFC 1321. DiffieHellman is the standard for the Diffie-Hellman key agreement algorithm defined in PKCS3.

In the JDK, there is an aliasing scheme that enables clients to use aliases when referring to algorithms or types, rather than their standard names. For example, the "SUN" provider's master class (Sun.java) defines the alias "SHA1/DSA" for the algorithm whose standard name is "SHA1withDSA". Thus, the following statements are equivalent: 

    Signature sig = Signature.getInstance("SHA1withDSA", "SUN");

    Signature sig = Signature.getInstance("SHA1/DSA", "SUN");

Aliases can be defined in your "master class" (see Step 3). To define an alias, create a property named

    Alg.Alias.engineClassName.aliasName

where engineClassName is the name of an engine class (e.g., Signature), and aliasName is your alias name. The value of the property must be the standard algorithm (or type) name for the algorithm (or type) being aliased.

As an example, the "SUN" provider defines the alias "SHA1/DSA" for the signature algorithm whose standard name is "SHA1withDSA" by setting a property named Alg.Alias.Signature.SHA1/DSA to have the value SHA1withDSA via the following: 

    put("Alg.Alias.Signature.SHA1/DSA", "SHA1withDSA");

Note that aliases defined by one provider are available only to that provider and not to any other providers. Thus, aliases defined by the SunJCE provider are available only to the SunJCE provider.

Service Interdependencies

Some algorithms require the use of other types of algorithms. For example, a PBE algorithm usually needs to use a message digest algorithm in order to transform a password into a key.

If you are implementing one type of algorithm that requires another, you can do one of the following:

  1. Provide your own implementations for both.
  2. Let your implementation of one algorithm use an instance of the other type of algorithm, as supplied by the default Sun provider that is included with every Java SE Platform installation. For example, if you are implementing a PBE algorithm that requires a message digest algorithm, you can obtain an instance of a class implementing the MD5 message digest algorithm by calling

        MessageDigest.getInstance("MD5", "SUN")
  3. Let your implementation of one algorithm use an instance of the other type of algorithm, as supplied by another specific provider. This is only appropriate if you are sure that all clients who will use your provider will also have the other provider installed.
  4. Let your implementation of one algorithm use an instance of the other type of algorithm, as supplied by another (unspecified) provider. That is, you can request an algorithm by name, but without specifying any particular provider, as in 
        MessageDigest.getInstance("MD5")
    This is only appropriate if you are sure that there will be at least one implementation of the requested algorithm (in this case, MD5) installed on each Java platform where your provider will be used.

Here are some common types of algorithm interdependencies:

Signature and Message Digest Algorithms

A signature algorithm often requires use of a message digest algorithm. For example, the SHA1withDSA signature algorithm requires the SHA-1 message digest algorithm.

Signature and (Pseudo-)Random Number Generation Algorithms

A signature algorithm often requires use of a (pseudo-)random number generation algorithm. For example, such an algorithm is required in order to generate a DSA signature.

Key Pair Generation and Message Digest Algorithms

A key pair generation algorithm often requires use of a message digest algorithm. For example, DSA keys are generated using the SHA-1 message digest algorithm.

Algorithm Parameter Generation and Message Digest Algorithms

An algorithm parameter generator often requires use of a message digest algorithm. For example, DSA parameters are generated using the SHA-1 message digest algorithm.

KeyStores and Message Digest Algorithms

A keystore implementation will often utilize a message digest algorithm to compute keyed hashes (where the key is a user-provided password) to check the integrity of a keystore and make sure that the keystore has not been tampered with.

Key Pair Generation Algorithms and Algorithm Parameter Generators

A key pair generation algorithm sometimes needs to generate a new set of algorithm parameters. It can either generate the parameters directly, or use an algorithm parameter generator.

Key Pair Generation, Algorithm Parameter Generation, and (Pseudo-)Random Number Generation Algorithms

A key pair generation algorithm may require a source of randomness in order to generate a new key pair and possibly a new set of parameters associated with the keys. That source of randomness is represented by a SecureRandom object. The implementation of the key pair generation algorithm may generate the key parameters itself, or may use an algorithm parameter generator to generate them, in which case it may or may not initialize the algorithm parameter generator with a source of randomness.

Algorithm Parameter Generators and Algorithm Parameters

An algorithm parameter generator's engineGenerateParameters method must return an AlgorithmParameters instance.

Signature and Key Pair Generation Algorithms or Key Factories

If you are implementing a signature algorithm, your implementation's engineInitSign and engineInitVerify methods will require passed-in keys that are valid for the underlying algorithm (e.g., DSA keys for the DSS algorithm). You can do one of the following:

  1. Also create your own classes implementing appropriate interfaces (e.g. classes implementing the DSAPrivateKey and DSAPublicKey interfaces from the package java.security.interfaces), and create your own key pair generator and/or key factory returning keys of those types. Require the keys passed to engineInitSign and engineInitVerify to be the types of keys you have implemented, that is, keys generated from your key pair generator or key factory. Or you can,

  2. Accept keys from other key pair generators or other key factories, as long as they are instances of appropriate interfaces that enable your signature implementation to obtain the information it needs (such as the private and public keys and the key parameters). For example, the engineInitSign method for a DSS Signature class could accept any private keys that are instances of java.security.interfaces.DSAPrivateKey.

KeyStores and Key and Certificate Factories

A keystore implementation will often utilize a key factory to parse the keys stored in the keystore, and a certificate factory to parse the certificates stored in the keystore.

Default Initializations

In case the client does not explicitly initialize a key pair generator or an algorithm parameter generator, each provider of such a service must supply (and document) a default initialization. For example, the Sun provider uses a default modulus size (strength) of 1024 bits for the generation of DSA parameters, and the "SunJCE" provider uses a default modulus size (keysize) of 1024 bits for the generation of Diffie-Hellman parameters.

Default Key Pair Generator Parameter Requirements

If you implement a key pair generator, your implementation should supply default parameters that are used when clients don't specify parameters. The documentation you supply (Step 11) should state what the default parameters are.

For example, the DSA key pair generator in the Sun provider supplies a set of pre-computed p, q, and g default values for the generation of 512, 768, and 1024-bit key pairs. The following p, q, and g values are used as the default values for the generation of 1024-bit DSA key pairs: 

p = fd7f5381 1d751229 52df4a9c 2eece4e7 f611b752 3cef4400 c31e3f80
    b6512669 455d4022 51fb593d 8d58fabf c5f5ba30 f6cb9b55 6cd7813b
    801d346f f26660b7 6b9950a5 a49f9fe8 047b1022 c24fbba9 d7feb7c6
    1bf83b57 e7c6a8a6 150f04fb 83f6d3c5 1ec30235 54135a16 9132f675
    f3ae2b61 d72aeff2 2203199d d14801c7

q = 9760508f 15230bcc b292b982 a2eb840b f0581cf5

g = f7e1a085 d69b3dde cbbcab5c 36b857b9 7994afbb fa3aea82 f9574c0b
    3d078267 5159578e bad4594f e6710710 8180b449 167123e8 4c281613
    b7cf0932 8cc8a6e1 3c167a8b 547c8d28 e0a3ae1e 2bb3a675 916ea37f
    0bfa2135 62f1fb62 7a01243b cca4f1be a8519089 a883dfe1 5ae59f06
    928b665e 807b5525 64014c3b fecf492a

(The p and q values given here were generated by the prime generation standard, using the 160-bit 

SEED:  8d515589 4229d5e6 89ee01e6 018a237e 2cae64cd

With this seed, the algorithm found p and q when the counter was at 92.)

The Provider.Service Class

Since its introduction, security providers have published their service information via appropriately formatted key-value String pairs they put in their Hashtable entries. While this mechanism is simple and convenient, it limits the amount customization possible. As a result, JDK 5.0 introduced a second option, the Provider.Service class. It offers an alternative way for providers to advertise their services and supports additional features as described below. Note that this addition is fully compatible with the older method of using String valued Hashtable entries. A provider on JDK 5.0 can choose either method as it prefers, or even use both at the same time.

A Provider.Service object encapsulates all information about a service. This is the provider that offers the service, its type (e.g. MessageDigest or Signature), the algorithm name, and the name of the class that implements the service. Optionally, it also includes a list of alternate algorithm names for this service (aliases) and attributes, which are a map of (name, value) String pairs. In addition, it defines the methods newInstance() and supportsParameter(). They have default implementations, but can be overridden by providers if needed, as may be the case with providers that interface with hardware security tokens.

The newInstance() method is used by the security framework when it needs to construct new implementation instances. The default implementation uses reflection to invoke the standard constructor for the respective type of service. For all standard services except CertStore, this is the no-args constructor. The constructorParameter to newInstance() must be null in theses cases. For services of type CertStore, the constructor that takes a CertStoreParameters object is invoked, and constructorParameter must be a non-null instance of CertStoreParameters. A security provider can override the newInstance() method to implement instantiation as appropriate for that implementation. It could use direct invocation or call a constructor that passes additional information specific to the Provider instance or token. For example, if multiple Smartcard readers are present on the system, it might pass information about which reader the newly created service is to be associated with. However, despite customization all implementations must follow the conventions about constructorParameter described above.

The supportsParameter() tests whether the Service can use the specified parameter. It returns false if this service cannot use the parameter. It returns true if this service can use the parameter, if a fast test is infeasible, or if the status is unknown. It is used by the security framework with some types of services to quickly exclude non-matching implementations from consideration. It is currently only defined for the following standard services: Signature, Cipher, Mac, and KeyAgreement. The parameter must be an instance of Key in these cases. For example, for Signature services, the framework tests whether the service can use the supplied Key before instantiating the service. The default implementation examines the attributes SupportedKeyFormats and SupportedKeyClasses as described below. Again, a provider may override this methods to implement additional tests.

The SupportedKeyFormats attribute is a list of the supported formats for encoded keys (as returned by key.getFormat()) separated by the "|" (pipe) character. For example, X.509|PKCS#8. The SupportedKeyClasses attribute is a list of the names of classes of interfaces separated by the "|" character. A key object is considered to be acceptable if it is assignable to at least one of those classes or interfaces named. In other words, if the class of the key object is a subclass of one of the listed classes (or the class itself) or if it implements the listed interface. An example value is "java.security.interfaces.RSAPrivateKey|java.security.interfaces.RSAPublicKey" .

Four methods have been added to the Provider class for adding and looking up Services. As mentioned earlier, the implementation of those methods and also of the existing Properties methods have been specifically designed to ensure compatibility with existing Provider subclasses. This is achieved as follows:

If legacy Properties methods are used to add entries, the Provider class makes sure that the property strings are parsed into equivalent Service objects prior to lookup via getService(). Similarly, if the putService() method is used, equivalent property strings are placed into the provider's hashtable at the same time. If a provider implementation overrides any of the methods in the Provider class, it has to ensure that its implementation does not interfere with this conversion. To avoid problems, we recommend that implementations do not override any of the methods in the Provider class.

Signature Formats

If you implement a signature algorithm, the documentation you supply (Step 11) should specify the format in which the signature (generated by one of the sign methods) is encoded.

For example, the SHA1withDSA signature algorithm supplied by the Sun provider encodes the signature as a standard ASN.1 sequence of two ASN.1 INTEGER values: r and s, in that order:

SEQUENCE ::= {
	r INTEGER,
	s INTEGER }

DSA Interfaces and their Required Implementations

The Java Security API contains the following interfaces (in the java.security.interfaces package) for the convenience of programmers implementing DSA services:

The following sections discuss requirements for implementations of these interfaces.

DSAKeyPairGenerator Implementation

This interface is obsolete. It used to be needed to enable clients to provide DSA-specific parameters to be used rather than the default parameters your implementation supplies. However, in Java it is no longer necessary; a new KeyPairGenerator initialize method that takes an AlgorithmParameterSpec parameter enables clients to indicate algorithm-specific parameters.

DSAParams Implementation

If you are implementing a DSA key pair generator, you need a class implementing DSAParams for holding and returning the p, q, and g parameters.

A DSAParams implementation is also required if you implement the DSAPrivateKey and DSAPublicKey interfaces. DSAPublicKey and DSAPrivateKey both extend the DSAKey interface, which contains a getParams method that must return a DSAParams object. See DSAPrivateKey and DSAPublicKey Implementations for more information.

Note: there is a DSAParams implementation built into the JDK: the java.security.spec.DSAParameterSpec class.

DSAPrivateKey and DSAPublicKey Implementations

If you implement a DSA key pair generator or key factory, you need to create classes implementing the DSAPrivateKey and DSAPublicKey interfaces.

If you implement a DSA key pair generator, your generateKeyPair method (in your KeyPairGeneratorSpi subclass) will return instances of your implementations of those interfaces.

If you implement a DSA key factory, your engineGeneratePrivate method (in your KeyFactorySpi subclass) will return an instance of your DSAPrivateKey implementation, and your engineGeneratePublic method will return an instance of your DSAPublicKey implementation.

Also, your engineGetKeySpec and engineTranslateKey methods will expect the passed-in key to be an instance of a DSAPrivateKey or DSAPublicKey implementation. The getParams method provided by the interface implementations is useful for obtaining and extracting the parameters from the keys and then using the parameters, for example as parameters to the DSAParameterSpec constructor called to create a parameter specification from parameter values that could be used to initialize a KeyPairGenerator object for DSA.

If you implement a DSA signature algorithm, your engineInitSign method (in your SignatureSpi subclass) will expect to be passed a DSAPrivateKey and your engineInitVerify method will expect to be passed a DSAPublicKey.

Please note: The DSAPublicKey and DSAPrivateKey interfaces define a very generic, provider-independent interface to DSA public and private keys, respectively. The engineGetKeySpec and engineTranslateKey methods (in your KeyFactorySpi subclass) could additionally check if the passed-in key is actually an instance of their provider's own implementation of DSAPrivateKey or DSAPublicKey, e.g., to take advantage of provider-specific implementation details. The same is true for the DSA signature algorithm engineInitSign and engineInitVerify methods (in your SignatureSpi subclass).

To see what methods need to be implemented by classes that implement the DSAPublicKey and DSAPrivateKey interfaces, first note the following interface signatures:

In the java.security.interfaces package:

   public interface DSAPrivateKey extends DSAKey,
		java.security.PrivateKey

   public interface DSAPublicKey extends DSAKey,
		java.security.PublicKey

   public interface DSAKey

In the java.security package:

   public interface PrivateKey extends Key

   public interface PublicKey extends Key

   public interface Key extends java.io.Serializable

In order to implement the DSAPrivateKey and DSAPublicKey interfaces, you must implement the methods they define as well as those defined by interfaces they extend, directly or indirectly.

Thus, for private keys, you need to supply a class that implements

  • the getX method from the DSAPrivateKey interface.

  • the getParams method from the java.security.interfaces.DSAKey interface, since DSAPrivateKey extends DSAKey. Note: The getParams method returns a DSAParams object, so you must also have a DSAParams implementation.

  • the getAlgorithm, getEncoded, and getFormat methods from the java.security.Key interface, since DSAPrivateKey extends java.security.PrivateKey, and PrivateKey extends Key.

      Similarly, for public DSA keys, you need to supply a class that implements:

    RSA Interfaces and their Required Implementations

    The Java Security API contains the following interfaces (in the java.security.interfaces package) for the convenience of programmers implementing RSA services:

    The following sections discuss requirements for implementations of these interfaces.

    RSAPrivateKey , RSAPrivateCrtKey , and RSAPublicKey Implementations

    If you implement an RSA key pair generator or key factory, you need to create classes implementing the RSAPrivateKey (and/or RSAPrivateCrtKey) and RSAPublicKey interfaces. (RSAPrivateCrtKey is the interface to an RSA private key, using the Chinese Remainder Theorem (CRT) representation.)

    If you implement an RSA key pair generator, your generateKeyPair method (in your KeyPairGeneratorSpi subclass) will return instances of your implementations of those interfaces.

    If you implement an RSA key factory, your engineGeneratePrivate method (in your KeyFactorySpi subclass) will return an instance of your RSAPrivateKey (or RSAPrivateCrtKey) implementation, and your engineGeneratePublic method will return an instance of your RSAPublicKey implementation.

    Also, your engineGetKeySpec and engineTranslateKey methods will expect the passed-in key to be an instance of an RSAPrivateKey, RSAPrivateCrtKey, or RSAPublicKey implementation.

    If you implement an RSA signature algorithm, your engineInitSign method (in your SignatureSpi subclass) will expect to be passed either an RSAPrivateKey or an RSAPrivateCrtKey, and your engineInitVerify method will expect to be passed an RSAPublicKey.

    Please note: The RSAPublicKey, RSAPrivateKey, and RSAPrivateCrtKey interfaces define a very generic, provider-independent interface to RSA public and private keys. The engineGetKeySpec and engineTranslateKey methods (in your KeyFactorySpi subclass) could additionally check if the passed-in key is actually an instance of their provider's own implementation of RSAPrivateKey, RSAPrivateCrtKey, or RSAPublicKey, e.g., to take advantage of provider-specific implementation details. The same is true for the RSA signature algorithm engineInitSign and engineInitVerify methods (in your SignatureSpi subclass).

    To see what methods need to be implemented by classes that implement the RSAPublicKey, RSAPrivateKey, and RSAPrivateCrtKey interfaces, first note the following interface signatures:

    In the java.security.interfaces package:

        public interface RSAPrivateKey extends java.security.PrivateKey

    public interface RSAPrivateCrtKey extends RSAPrivateKey

    public interface RSAPublicKey extends java.security.PublicKey

    In the java.security package:

        public interface PrivateKey extends Key

    public interface PublicKey extends Key

    public interface Key extends java.io.Serializable

    In order to implement the RSAPrivateKey, RSAPrivateCrtKey, and RSAPublicKey interfaces, you must implement the methods they define as well as those defined by interfaces they extend, directly or indirectly.

    Thus, for RSA private keys, you need to supply a class that implements:

    Similarly, for RSA private keys using the Chinese Remainder Theorem (CRT) representation, yo