Java Security by Design

Application security is critical in Java technology. The Java runtime environment (aka JVM) provides a tightly guarded security environment for runtime execution. (Refer to [J2EE14], [J2EE14Tutor], and [LiGong] for more details.) Figure 13-1 depicts a high-level security overview. Inside the JRE, the Security Manager is responsible for code runtime verification and access control. The code runtime verification is managed by the Protection Domain, where different class files (namely, bootstrap class, system class, and user class) are verified by bootstrap class loader, system class loader, class loader, and the bytecode verifier. The Access Module is responsible to authenticate and authorize the principal (user or service requester) against the security policy files (namely, java.security and java.policy files). The JRE supports a variety of authentication mechanisms, including JAAS login module, database security (using JDBC), or LDAP (using JNDI).

Figure 13-1. Java security overview

Java Runtime Security

Java SE provides a Java "sandbox" to restrict applets from accessing file systems and networks and untrusted applets from accessing all of the API functionality. The "sandbox" security architecture (refer to Figure 13-1) consists of three key components:

• Bytecode Verifier It verifies Java bytecodes that are compiled from Java source codes prior to execution. It ensures that the bytecodes do not violate permission policies or access system resources using incorrect type information.

• Class Loader The "primordial" (root) class loader bootstraps the class loading process and protects the runtime environment from loading harmful codes locally or remotely by hackers during the class loading process (so-called "code-spoofing"). It loads the initial classes required by all Java programs. The secure class loader java.security.SecureClassLoader then kicks off assembling and loading other classes locally, for example, bootstrap class files, system class files, and user class files.

  Class loading works under a class loader hierarchy. A child class must delegate to the parent class to load a specific class. If the parent class cannot load the specific class, then the child class loads it. Remote classes from the network are instantiated and loaded by the parent class loader as a new class. Thus hackers are not able to spoof attack by loading a malicious class directly into the JVM. For instance, hackers are able to insert a malicious version of System.out.print or java.lang.String into the application because the class loader loads a local version of System.out.print or java.lang.String under this class loading process. A Java archive (JAR) signer is a utility that seals packages for protection from tampering and verifies remote classes prior to loading to the JVM.

• Security Manager The security manager (java.lang.SecurityManager) performs runtime checks on any method or any code accessing sensitive system resources (for example, file or network access), and generates a security exception for any security policy violation. It delegates the permission check to the java.security.AccessController by calling the checkPermission method. The security manager can be invoked by specifying the system property while starting the JVM (for example, java -Djava.security.manager myApps.class) or creating an instance in the program code (for example, System.setSecurityManager(new SecurityManager());)

The security manager has a security policy database where security policies are maintained. The security policy database stores permission rules for authorization and key stores for authentication. A security policy relates to a set of permissions for a domain (system or application), which encloses a set of classes. Developers can also customize any additional protection of resources within the domain boundary, say, using the SignedObject class.

The protection domain (java.security.ProtectionDomain class) is another important security concept. It refers to the system and application components (for example, a group of classes) of the runtime environment that can be secured according to the predefined security policy. There are two types of protection domain: static (grant permissions specified only when constructed) and dynamic (grant permissions specified when constructed and permissions granted by the security policy). The protection domain extends the "sandbox" security architecture by associating a group of principals with permissions. For example, a protection domain associates permissions with a code source (URL where the class file comes from), such that any classes originating from the same URL will have the same signature and key placed in the same domain, and thus granted the same permissions and access rights. This enhances the current mechanism to load a class. Moreover, the security policy allows defining an association between the principals and permissions for the classes (via the code source). It can be passed as a parameter to the application, which can use different protection domains wherever necessary.

Java SE security introduces a set of additional security concepts (see [LiGong] for details), which includes the concepts of a Principal (an entity that a security service can authenticate with an authentication protocol) and security domain (the scope related to a set of security policies defined by the administrator of a security service). There are two important files under JVM that store security policy information: java.policy and java.security under the directory, %J2SE%/lib/security, where %J2SE% is the file location for the JVM.

Authentication Mechanisms

Figure 13-1 shows an example of three different authentication mechanisms. A subject (user) has multiple principals, or multiple user names or identities. Suppose one of his principals (using the java.security.Principal class) possesses a digital certificate as a credential. Under the Java SE security architecture, the principal can use the credential to authenticate with the applications via JAAS (using the javax.security.auth.login.* class), JNDI (using the LDAP directory server), or JDBC (using back-end database security). In this example, the security manager is used. Upon successful authentication, the security manager will check permission, and pass control to the access module. The access module (using the java.security.AccessController class) checks permission by checking the java.security file, which contains the policy URL (policy.url.1) and keystore information (keystore.type). The java.security file associates the permission (grant ... permission ...) with the principals. If the principal-permission relationship is found, then the access module grants access to the principal for the application resources to the principal.

Apart from the "sandbox" security architecture, Java SE also provides some authentication and encryption services that work with JCA and JCE layers. These security services include JAAS, JSSE, JGSS, and CertPath. Applications or security packages can also customize their security APIs using these security services.

Container-Based Security

The Java EE container provides a comprehensive application-level security that is related to the application component deployment and runtime environment.

• Declarative Security According to [J2EE14], Java EE security has the notion of declarative security. In other words, the application's security structure, including security roles, access control, and authentication requirements in a form external to the application, is expressed in the deployment descriptor. The deployment descriptor is in effect a contract between the application component provider and the deployer or application assembler, where the application security policy is mapped to the security structure of the relevant operating environment.

• Programmatic Security Java EE security architecture also provides some APIs (programmatic security) to manipulate the roles and principals, in addition to the declarative security. This supports both servlets (isUserInRole and getUserPrincipal using the interface HttpServletRequest) and EJBs (isCallerInRole and getCallerPrincipal using the interface EJBContext).

Security Interoperability Features

Java technology has provided several interoperability features to secure business applications. The following highlights a few major security interoperability features:

• "Building block" security components to support interoperability, for example, JAAS and JSSE.

• JSR implementation that enables interoperability, for example, JAX-RPC 2.0 and WS-I Basic Security Profile 1.0/1.1.

• Support of single sign-on using Web SSO protocol, WS-MEX protocol, Liberty, and SAML.

• Support of security interoperability standards, for example, OASIS's Web services security.

• Support of WS-Policy by Java Web Services Developer Pack 2.1 or later.