Spring Security provides you with a flexible framework for your authentication and authorization requirements, but there are many other considerations for building a secure application that are outside its scope.
Web applications are vulnerable to all kinds of attacks with which you should be familiar, preferably before you start development so that you can design and code with them in mind from the beginning.
Check out the https://www.owasp.org/[OWASP web site] for information on the major issues that face web application developers and the countermeasures you can use against them.
Suppose you are developing an enterprise application based on Spring.
You typically need to address four security concerns : authentication, web request security, service layer security (your methods that implement business logic), and domain object instance security (different domain objects can have different permissions). With these typical requirements in mind, we have the following considerations:
* _Authentication_: The servlet specification provides an approach to authentication.
However, you need to configure the container to perform authentication, which typically requires editing of container-specific "`realm`" settings.
This makes a non-portable configuration. Also, if you need to write an actual Java class to implement the container's authentication interface, it becomes even more non-portable.
With Spring Security, you achieve complete portability -- right down to the WAR level.
Also, Spring Security offers a choice of production-proven authentication providers and mechanisms, meaning you can switch your authentication approaches at deployment time.
This is particularly valuable for software vendors writing products that need to work in an unknown target environment.
* _Service layer and domain object security:_ The absence of support in the servlet specification for services layer security or domain object instance security represents serious limitations for multi-tiered applications.
Typically, developers either ignore these requirements or implement security logic within their MVC controller code (or, even worse, inside the views). There are serious disadvantages with this approach:
** _Separation of concerns:_ Authorization is a crosscutting concern and should be implemented as such.
MVC controllers or views that implement authorization code makes it more difficult to test both the controller and the authorization logic, is more difficult to debug, and often leads to code duplication.
** _Support for rich clients and web services:_ If an additional client type must ultimately be supported, any authorization code embedded within the web layer is non-reusable.
It should be considered that Spring remoting exporters export only service layer beans (not MVC controllers). As a result, authorization logic needs to be located in the services layer to support a multitude of client types.
** _Layering issues:_ An MVC controller or view is the incorrect architectural layer in which to implement authorization decisions concerning services layer methods or domain object instances.
While the principal may be passed to the services layer to enable it to make the authorization decision, doing so would introduce an additional argument on every services layer method.
A more elegant approach is to use a `ThreadLocal` to hold the principal, although this would likely increase development time to a point where it would become more economical (on a cost-benefit basis) to use a dedicated security framework.
** _Authorisation code quality:_ It is often said of web frameworks that they "`make it easier to do the right things, and harder to do the wrong things`". Security frameworks are the same, because they are designed in an abstract manner for a wide range of purposes.
Writing your own authorization code from scratch does not provide the "`design check`" a framework would offer, and in-house authorization code typically lacks the improvements that emerge from widespread deployment, peer review, and new versions.
Although when considered within the context of web container portability, configuration requirements, limited web request security flexibility, and non-existent services layer and domain object instance security, it becomes clear why developers often look to alternative solutions.
Suppose you are new to Spring Security and need to build an application that supports CAS single sign-on over HTTPS while allowing basic authentication locally for certain URLs, authenticating against multiple back end user information sources (LDAP and JDBC). You have copied some configuration files but have found that it does not work. What could be wrong?
If you use other technologies with which you are not familiar, you should do some research and try to make sure you can use them in isolation before combining them in a complex system.
This also means that, if you ask this question online, you should not expect an answer unless you provide additional information.
As with any issue, you should check the output from the debug log and note any exception stacktraces and related messages.
You should step through the code in a debugger to see where the authentication fails and why.
You should also write a test case which exercises your authentication configuration outside of the application.
If you use hashed passwords, make sure the value stored in your database is _exactly_ the same as the value produced by the `PasswordEncoder` configured in your application.
A common user problem with infinite loop and redirecting to the login page is caused by accidentally configuring the login page as a "`secured`" resource.
Make sure your configuration allows anonymous access to the login page, either by excluding it from the security filter chain or marking it as requiring `ROLE_ANONYMOUS`.
If your `AccessDecisionManager` includes an `AuthenticatedVoter`, you can use the `IS_AUTHENTICATED_ANONYMOUSLY` attribute. This is automatically available if you use the standard namespace configuration setup.
From Spring Security 2.0.1 onwards, when you use namespace-based configuration, a check is made on loading the application context and a warning message logged if your login page appears to be protected.
The most common reason for this is that your browser has cached the page and you are seeing a copy that is being retrieved from the browsers cache.
Verify this by checking whether the browser is actually sending the request (check your server access logs and the debug log or use a suitable browser debugging plugin, such as "`Tamper Data`" for Firefox). This has nothing to do with Spring Security, and you should configure your application or server to set the appropriate `Cache-Control` response headers.
The following listing shows another debug-level message that occurs the first time an anonymous user attempts to access a protected resource. However, this listing shows what happens when you do not have an `AnonymousAuthenticationFilter` in your filter chain configuration:
Note that the permissions for an LDAP directory often do not let you read the password for a user.
Hence, it is often not possible to use the <<appendix-faq-what-is-userdetailservice>> where Spring Security compares the stored password with the one submitted by the user.
The most common approach is to use LDAP "`bind`", which is one of the operations supported by https://en.wikipedia.org/wiki/Lightweight_Directory_Access_Protocol[the LDAP protocol]. With this approach, Spring Security validates the password by trying to authenticate to the directory as the user.
This differs from one company to another, so you have to find it out yourself.
Before adding a Spring Security LDAP configuration to an application, you should write a simple test by using standard Java LDAP code (without Spring Security involved) and make sure you can get that to work first.
You should also understand the difference between secure and non-secure cookies and the implications of using HTTP and HTTPS and switching between the two.
=== I am using Spring Security's concurrent session control to prevent users from logging in more than once at the same time. When I open another browser window after logging in, it does not stop me from logging in again. Why can I log in more than once?
So, if you log in again in another window or tab, you are reauthenticating in the same session.
The server does not know anything about tabs, windows, or browser instances.
All it sees are HTTP requests, and it ties those to a particular session according to the value of the `JSESSIONID` cookie that they contain.
When a user authenticates during a session, Spring Security's concurrent session control checks the number of _other authenticated sessions_ that they have.
If they are already authenticated with the same session, re-authenticating has no effect.
If you us a Servlet 3.1 or newer container, the session ID is simply changed.
If you use an older container, Spring Security invalidates the existing session, creates a new session, and transfers the session data to the new session.
Changing the session identifier in this manner prevents "`session-fixation`" attacks.
=== I use Tomcat (or some other servlet container) and have enabled HTTPS for my login page, switching back to HTTP afterwards. It does not work. I end up back at the login page after authenticating.
This happens because sessions created under HTTPS, for which the session cookie is marked as "`secure`", cannot subsequently be used under HTTP. The browser does not send the cookie back to the server, and any session state (including the security context information) is lost. Starting a session in HTTP first should work, as the session cookie is not marked as secure.
However, Spring Security's https://docs.spring.io/spring-security/site/docs/3.1.x/reference/springsecurity-single.html#ns-session-fixation[Session Fixation Protection] can interfere with this because it results in a new session ID cookie being sent back to the user's browser, usually with the secure flag.
To get around this, you can disable session fixation protection. However, in newer Servlet containers, you can also configure session cookies to never use the secure flag.
Sessions are maintained either by exchanging a session cookie or by adding a `jsessionid` parameter to URLs (this happens automatically if you use JSTL to output URLs or if you call `HttpServletResponse.encodeUrl` on URLs (before a redirect, for example). If clients have cookies disabled and you are not rewriting URLs to include the `jsessionid`, the session is lost.
=== I am trying to use the concurrent session-control support, but it does not let me log back in, even if I am sure I have logged out and have not exceeded the allowed sessions. What is wrong?
=== Spring Security creates a session somewhere, even though I have configured it not to, by setting the create-session attribute to never. What is wrong?
If you have trouble working out where a session is being created, you can add some debugging code to track down the location(s). One way to do this is to add a `javax.servlet.http.HttpSessionListener`, which calls `Thread.dumpStack()` in the `sessionCreated` method, to your application.
If an HTTP 403 Forbidden error is returned for HTTP POST but it works for HTTP GET, the issue is most likely related to https://docs.spring.io/spring-security/site/docs/3.2.x/reference/htmlsingle/#csrf[CSRF]. Either provide the CSRF Token or disable CSRF protection (the latter is not recommended).
By default, filters are not applied to forwards or includes.
If you really want the security filters to be applied to forwards or includes, you have to configure these explicitly in your `web.xml` file by using the `<dispatcher>` element, which is a child element of the `<filter-mapping>` element.
=== I have added Spring Security's <global-method-security> element to my application context, but, if I add security annotations to my Spring MVC controller beans (Struts actions etc.), they do not seem to have an effect. Why not?
In a Spring web application, the application context that holds the Spring MVC beans for the dispatcher servlet is often separate from the main application context.
It is often defined in a file called `myapp-servlet.xml`, where `myapp` is the name assigned to the Spring `DispatcherServlet` in the `web.xml` file. An application can have multiple `DispatcherServlet` instances, each with its own isolated application context.
The beans in these "`child`" contexts are not visible to the rest of the application.
The "`parent`" application context is loaded by the `ContextLoaderListener` you define in your `web.xml` file and is visible to all the child contexts.
This parent context is usually where you define your security configuration, including the `<global-method-security>` element. As a result, any security constraints applied to methods in these web beans are not enforced, since the beans cannot be seen from the `DispatcherServlet` context.
You need to either move the `<global-method-security>` declaration to the web context or move the beans you want secured into the main application context.
=== I have a user who has definitely been authenticated, but, when I try to access the SecurityContextHolder during some requests, the Authentication is null. Why can I not see the user information?
If you have excluded the request from the security filter chain by using the `filters='none'` attribute in the `<intercept-url>` element that matches the URL pattern, the `SecurityContextHolder` is not populated for that request.
Method security does not hide links when using the `url` attribute in `<sec:authorize>`, because we cannot readily reverse engineer what URL is mapped to what controller endpoint. We are limited because controllers can rely on headers, the current user, and other details to determine what method to invoke.
The best way of locating classes is by installing the Spring Security source in your IDE. The distribution includes source jars for each of the modules the project is divided up into.
Add these to your project source path and you can navigate directly to Spring Security classes (`Ctrl-Shift-T` in Eclipse). This also makes debugging easier and lets you troubleshoot exceptions by looking directly at the code where they occur to see what is going on there.
There is also a detailed blog article called "Behind the Spring Security Namespace" on https://spring.io/blog/2010/03/06/behind-the-spring-security-namespace/[blog.springsource.com]. If want to know the full details then the code is in the `spring-security-config` module within the Spring Security 3.0 distribution.
Spring Security has a voter-based architecture, which means that an access decision is made by a series of `AccessDecisionVoter` instances.
The voters act on the "`configuration attributes`", which are specified for a secured resource (such as a method invocation). With this approach, not all attributes may be relevant to all voters, and a voter needs to know when it should ignore an attribute (abstain) and when it should vote to grant or deny access based on the attribute value.
The most common voter is the `RoleVoter`, which, by default, votes whenever it finds an attribute with the `ROLE_` prefix.
It makes a simple comparison of the attribute (such as `ROLE_USER`) with the names of the authorities that the current user has been assigned.
If it finds a match (they have an authority called `ROLE_USER`), it votes to grant access. Otherwise, it votes to deny access.
You can change the prefix by setting the `rolePrefix` property of `RoleVoter`. If you need only to use roles in your application and have no need for other custom voters, you can set the prefix to a blank string. In that case, the `RoleVoter` treats all attributes as roles.
With Spring Security 3.0, the project jars are divided into clearly distinct areas of functionality, so it is straightforward to work out which Spring Security jars you need from your application requirements.
All applications need the `spring-security-core` jar.
If you are developing a web application, you need the `spring-security-web` jar.
If you are using security namespace configuration, you need the `spring-security-config` jar. For LDAP support, you need the `spring-security-ldap` jar. And so on.
If you want to use LDAP with an embedded test server, use the LDAP sample as a starting point.
The reference manual also includes <<appendix-namespace,an appendix>> that lists the first-level dependencies for each Spring Security module, with some information on whether they are optional and when they are required.
If you build your project with Maven, adding the appropriate Spring Security modules as dependencies to your `pom.xml` file automatically pulls in the core jars that the framework requires.
Any that are marked as "`optional`" in the Spring Security `pom.xml` files have to be added to your own `pom.xml` file if you need them.
Authenticating a user with a username and password combination is most commonly performed by the `DaoAuthenticationProvider`, which is injected with a `UserDetailsService` to let it to load the password (and other data) for a user, to compare it with the submitted value.
Note that, if you use LDAP, <<appendix-faq-ldap-authentication,this approach may not work>>.
If you want to customize the authentication process, you should implement `AuthenticationProvider` yourself.
See this https://spring.io/blog/2010/08/02/spring-security-in-google-app-engine/[ blog article] for an example that integrate Spring Security authentication with Google App Engine.
The submitted login information is processed by an instance of `UsernamePasswordAuthenticationFilter`. You need to customize this class to handle the extra data fields. One option is to use your own customized authentication token class (rather than the standard `UsernamePasswordAuthenticationToken`). Another option is to concatenate the extra fields with the username (for example, by using a `:` character as the separator) and pass them in the username property of `UsernamePasswordAuthenticationToken`.
You also need to customize the actual authentication process.
If you use a custom authentication token class, for example, you will have to write an `AuthenticationProvider` (or extend the standard `DaoAuthenticationProvider`) to handle it. If you have concatenated the fields, you can implement your own `UserDetailsService` to split them up and load the appropriate user data for authentication.
=== How do I apply different intercept-url constraints where only the fragment value of the requested URLs differs (such as /thing1#thing2 and /thing1#thing3?
Instead of implementing `UserDetailsService`, you should implement `AuthenticationProvider` directly and extract the information from the supplied `Authentication` token.
In a standard web setup, the `getDetails()` method on the `Authentication` object will return an instance of `WebAuthenticationDetails`. If you need additional information, you can inject a custom `AuthenticationDetailsSource` into the authentication filter you are using.
If you are using the namespace, for example with the `<form-login>` element, then you should remove this element and replace it with a `<custom-filter>` declaration pointing to an explicitly configured `UsernamePasswordAuthenticationFilter`.
You cannot, since the `UserDetailsService` has no awareness of the servlet API. If you want to store custom user data, you should customize the `UserDetails` object that is returned.
This can then be accessed at any point, through the thread-local `SecurityContextHolder`. A call to `SecurityContextHolder.getContext().getAuthentication().getPrincipal()` returns this custom object.
People often ask about how to store the mapping between secured URLs and security metadata attributes in a database rather than in the application context.
A security-conscious organization should be aware that the benefits of their diligent testing process could be wiped out instantly by letting the security settings be modified at runtime by changing a row or two in a configuration database.
If you have taken this into account (perhaps by using multiple layers of security within your application), Spring Security lets you fully customize the source of security metadata.
Both method and web security are protected by subclasses of `AbstractSecurityInterceptor`, which is configured with a `SecurityMetadataSource` from which it obtains the metadata for a particular method or filter invocation.
For web security, the interceptor class is `FilterSecurityInterceptor`, and it uses the `FilterInvocationSecurityMetadataSource` marker interface. The "`secured object`" type it operates on is a `FilterInvocation`. The default implementation (which is used both in the namespace `<http>` and when configuring the interceptor explicitly) stores the list of URL patterns and their corresponding list of "`configuration attributes`" (instances of `ConfigAttribute`) in an in-memory map.
To load the data from an alternative source, you must use an explicitly declared security filter chain (typically Spring Security's `FilterChainProxy`) to customize the `FilterSecurityInterceptor` bean.
You cannot use the namespace.
You would then implement `FilterInvocationSecurityMetadataSource` to load the data as you please for a particular `FilterInvocation`. The `FilterInvocation` object contains the `HttpServletRequest`, so you can obtain the URL or any other relevant information on which to base your decision, based on what the list of returned attributes contains. A basic outline would look something like the following example:
The `LdapAuthenticationProvider` bean (which handles normal LDAP authentication in Spring Security) is configured with two separate strategy interfaces, one that performs the authentication and one that loads the user authorities, called `LdapAuthenticator` and `LdapAuthoritiesPopulator`, respectively.
The `DefaultLdapAuthoritiesPopulator` loads the user authorities from the LDAP directory and has various configuration parameters to let you specify how these should be retrieved.
You would then add a bean of this type to your application context and inject it into the `LdapAuthenticationProvider`. This is covered in the section on configuring LDAP by using explicit Spring beans in the LDAP chapter of the reference manual.
Note that you cannot use the namespace for configuration in this case.
You should also consult the security-api-url[Javadoc] for the relevant classes and interfaces.
=== I want to modify the property of a bean that is created by the namespace, but there is nothing in the schema to support it. What can I do short of abandoning namespace use?
The namespace functionality is intentionally limited, so it does not cover everything that you can do with plain beans.
If you want to do something simple, such as modifying a bean or injecting a different dependency, you can do so by adding a `BeanPostProcessor` to your configuration.
You can find more information in the https://docs.spring.io/spring/docs/3.0.x/spring-framework-reference/htmlsingle/spring-framework-reference.html#beans-factory-extension-bpp[Spring Reference Manual]. To do so, you need to know a bit about which beans are created, so you should also read the blog article mentioned in the earlier question on <<appendix-faq-namespace-to-bean-mapping,how the namespace maps to Spring beans>>.
Normally, you would add the functionality you require to the `postProcessBeforeInitialization` method of `BeanPostProcessor`. Suppose that you want to customize the `AuthenticationDetailsSource` used by the `UsernamePasswordAuthenticationFilter` (created by the `form-login` element). You want to extract a particular header called `CUSTOM_HEADER` from the request and use it while authenticating the user.
The processor class would look like the following listing:
