I am trying to implements additional checks for a user which is exchanging code for tokens using "/oauth2/token" endpoint in Spring Authorization Server. And for this I need to provide custom error message, error code and provide specific http status(other than 400 or 500).
I see that the code exchange starts in OAuth2TokenEndpointFilter but it has a strict exception hanling like
private void sendErrorResponse(HttpServletRequest request, HttpServletResponse response,
AuthenticationException exception) throws IOException { ... }
and it can not be overridden as well as can not be set
private AuthenticationFailureHandler authenticationFailureHandler = this::sendErrorResponse;
So I can extend from OAuth2AuthenticationException but it does not suite as I can not control the status and the response body.
Ok, I should read doc more carefully.
I still have to extend from AuthenticationException but I also have full controll over failure so adding custom body/code/status
#Bean
public SecurityFilterChain authorizationServerSecurityFilterChain(HttpSecurity http) throws Exception {
OAuth2AuthorizationServerConfigurer authorizationServerConfigurer =
new OAuth2AuthorizationServerConfigurer();
http.apply(authorizationServerConfigurer);
authorizationServerConfigurer
.tokenEndpoint(tokenEndpoint ->
((OAuth2TokenEndpointConfigurer)tokenEndpoint).errorResponseHandler(errorResponseHandler)// instance of AuthenticationFailureHandler
);
return http.build();
}
Related
I'm trying to build an Identity Provider using Spring authorization-server that third party applications are going to use for FIM (federated identity management).
We want each OAuth client to require authentication (if a user tries to login with a different client they would need to authenticate for each client).
Out of the box the flow looks like this:
So there's 2 issues.
The /oauth2/authorize endpoint just checks whether or not the sessions principal is authenticated, it doesn't care or know which client the principal was meant for.
There's just a single /login endpoint, so during authentication it doesn't know which client is used.
My best bet here is that I should:
Make the oauth2/authorize endpoint redirection to /login include the query parameter client_id
Create a custom AuthenticationFilter that also adds the client_id to the User principal
Override the authorizationRequestConverter for the oauth2/authorize endpoint and validate that the client in the request is the same as the client stored on the authenticated principal
Am I missing anything or do anyone know of a simpler way of doing this?
Based on your last comment, it seems one possibility is to simply require authentication every time, or at least every time an authorization is requested. In that case, you could clear out the authentication after the authorization code is issued to the client, using a Filter. This doesn't seem ideal and will result in a poor user experience, but may achieve your requirement.
#Configuration
#EnableWebSecurity
public class SecurityConfig {
#Bean
#Order(1)
public SecurityFilterChain authorizationServerSecurityFilterChain(HttpSecurity http)
throws Exception {
OAuth2AuthorizationServerConfiguration.applyDefaultSecurity(http);
// ...
// Add filter to remove the SecurityContext after successful authorization
http.addFilterAfter(new RemoveSecurityContextOnAuthorizationFilter(), LogoutFilter.class);
return http.build();
}
private static final class RemoveSecurityContextOnAuthorizationFilter extends OncePerRequestFilter {
private SecurityContextHolderStrategy securityContextHolderStrategy =
SecurityContextHolder.getContextHolderStrategy();
private final LogoutHandler logoutHandler = new CompositeLogoutHandler(
new CookieClearingLogoutHandler("JSESSIONID"),
new SecurityContextLogoutHandler()
);
#Override
protected void doFilterInternal(HttpServletRequest request, HttpServletResponse response, FilterChain filterChain) throws ServletException, IOException {
try {
filterChain.doFilter(request, response);
} finally {
String locationHeader = response.getHeader(HttpHeaders.LOCATION);
if (locationHeader != null) {
UriComponents uriComponents = UriComponentsBuilder.fromUriString(locationHeader).build();
if (uriComponents.getQueryParams().containsKey("code")) {
Authentication authentication = this.securityContextHolderStrategy.getContext().getAuthentication();
this.logoutHandler.logout(request, response, authentication);
}
}
}
}
}
// ...
}
I have a vaadin14 application that I want to enable different types of authentication mechanisms on different url paths. One is a test url, where authentication should use DB, and the other is the production url that uses keycloak.
I was able to get each authentication mechanism to work separately, but once I try to put both, I get unexpected results.
In both cases, I get login page, but the authentication doesn't work correctly. Here's my security configuration, what am I doing wrong?
#Configuration
#EnableWebSecurity
public class ApplicationSecurityConfiguration {
#Configuration
#Order(2)
public static class DBAuthConfigurationAdapter extends WebSecurityConfigurerAdapter {
private static final String LOGIN_PROCESSING_URL = "/login";
private static final String LOGIN_FAILURE_URL = "/login?error";
private static final String LOGIN_URL = "/login";
private static final String LOGOUT_SUCCESS_URL = "/login";
/**
* Require login to access internal pages and configure login form.
*/
#Override
protected void configure(HttpSecurity http) throws Exception {
// Not using Spring CSRF here to be able to use plain HTML for the login page
http.csrf().disable()
// Register our CustomRequestCache, that saves unauthorized access attempts, so
// the user is redirected after login.
.requestCache().requestCache(new CustomRequestCache())
// Restrict access to our application.
.and().antMatcher("/test**").authorizeRequests()
// Allow all flow internal requests.
.requestMatchers(SecurityUtils::isFrameworkInternalRequest).permitAll()
// Allow all requests by logged in users.
.anyRequest().hasRole("USER")
// Configure the login page.
.and().formLogin().loginPage(LOGIN_URL).permitAll().loginProcessingUrl(LOGIN_PROCESSING_URL)
.failureUrl(LOGIN_FAILURE_URL)
// Configure logout
.and().logout().logoutSuccessUrl(LOGOUT_SUCCESS_URL);
}
#Bean
#Override
public UserDetailsService userDetailsService() {
Properties users = null;
try {
users = PropertiesLoaderUtils.loadAllProperties("users.properties");
return new InMemoryUserDetailsManager(users);
} catch (IOException e) {
e.printStackTrace();
}
UserDetails user =
User.withUsername("user")
.password("{noop}password")
.roles("ACTOR")
.build();
return new InMemoryUserDetailsManager(user);
}
/**
* Allows access to static resources, bypassing Spring security.
*/
#Override
public void configure(WebSecurity web) {
web.ignoring().antMatchers(
// Vaadin Flow static resources
"/VAADIN/**",
// the standard favicon URI
"/favicon.ico",
// the robots exclusion standard
"/robots.txt",
// web application manifest
"/manifest.webmanifest",
"/sw.js",
"/offline-page.html",
// icons and images
"/icons/**",
"/images/**",
// (development mode) static resources
"/frontend/**",
// (development mode) webjars
"/webjars/**",
// (development mode) H2 debugging console
"/h2-console/**",
// (production mode) static resources
"/frontend-es5/**", "/frontend-es6/**",
"/resources/**");
}
}
#Order(1)
#Configuration
#ComponentScan(basePackageClasses = KeycloakSecurityComponents.class)
public static class AppKeycloakSecurity extends KeycloakWebSecurityConfigurerAdapter {
#Autowired
public void configureGlobal(
AuthenticationManagerBuilder auth) throws Exception {
KeycloakAuthenticationProvider keycloakAuthenticationProvider
= keycloakAuthenticationProvider();
keycloakAuthenticationProvider.setGrantedAuthoritiesMapper(
new SimpleAuthorityMapper());
auth.authenticationProvider(keycloakAuthenticationProvider);
}
#Bean
public KeycloakSpringBootConfigResolver KeycloakConfigResolver() {
return new KeycloakSpringBootConfigResolver();
}
#Override
protected SessionAuthenticationStrategy sessionAuthenticationStrategy() {
return new RegisterSessionAuthenticationStrategy(
new SessionRegistryImpl());
}
#Override
protected void configure(HttpSecurity http) throws Exception {
super.configure(http);
http.httpBasic().disable();
http.formLogin().disable();
http.anonymous().disable();
http.csrf().disable();
http.headers().frameOptions().disable();
http
.antMatcher("/prod**")
.authorizeRequests()
.antMatchers("/vaadinServlet/UIDL/**").permitAll()
.antMatchers("/vaadinServlet/HEARTBEAT/**").permitAll()
.requestMatchers(SecurityUtils::isFrameworkInternalRequest).permitAll()
.anyRequest().hasRole("actor");
http
.logout()
.addLogoutHandler(keycloakLogoutHandler())
.logoutUrl("/logout").permitAll()
.logoutSuccessUrl("/");
http
.addFilterBefore(keycloakPreAuthActionsFilter(), LogoutFilter.class);
http
.exceptionHandling()
.authenticationEntryPoint(authenticationEntryPoint());
http
.sessionManagement()
.sessionAuthenticationStrategy(sessionAuthenticationStrategy());
}
}
}
Navigating within a Vaadin UI will change the URL in your browser, but it will not necessarily create a browser request to that exact URL, effectively bypassing the access control defined by Spring security for that URL. As such, Vaadin is really not suited for the request URL-based security approach that Spring provides. For this issue alone you could take a look at my add-on Spring Boot Security for Vaadin which I specifically created to close the gap between Spring security and Vaadin.
But while creating two distinct Spring security contexts based on the URL is fairly easy, this - for the same reason - will not work well or at all with Vaadin. And that's something even my add-on couldn't help with.
Update: As combining both security contexts is an option for you, I can offer the following solution (using my add-on):
Starting from the Keycloak example, you would have to do the following:
Change WebSecurityConfig to also add your DB-based AuthenticationProvider. Adding your UserDetailsService should still be enough. Make sure to give every user a suitable role.
You have to remove this line from application.properties: codecamp.vaadin.security.standard-auth.enabled = false
This will re-enable the standard login without Keycloak via a Vaadin view.
Adapt the KeycloakRouteAccessDeniedHandler to ignore all test views that shouldn't be protected by Keycloak.
I already prepared all this in Gitlab repo and removed everything not important for the main point of this solution. See the individual commits and their diffs to also help focus in on the important bits.
I use this statements in the WebSecurityConfigurerAdapter#configure method to secure my application.
If csrf is disabled (http.csrf().disable();) than everything works fine, if I remove this line than I get an exception (Forbidden 403).
It is not clear for me why - I want to be safe against csrf- attack.
Does anyone know what I am doing wrong?
#Override
protected void configure(final HttpSecurity http) throws Exception {
final List<String> activeProfiles = Arrays.asList(environment.getActiveProfiles());
final CsrfTokenResponseHeaderBindingFilter csrfFilter = csrfTokenResponseHeaderBindingFilter();
http.addFilterAfter(csrfFilter, CsrfFilter.class).headers().cacheControl().xssProtection();
http.csrf().disable();
http.sessionManagement().sessionCreationPolicy(SessionCreationPolicy.STATELESS);
final String[] restEndpointsToSecure = WebSecurityConfig.restEndpointsToSecure;
for (final String endpoint : restEndpointsToSecure) {
http.authorizeRequests().antMatchers("/" + endpoint + "/**").hasRole(UserRoleEnum.USER.toString());
}
xAuthTokenConfigurer.setDetailsService(userDetailsServiceBean());
final SecurityConfigurer<DefaultSecurityFilterChain, HttpSecurity> securityConfigurerAdapter = xAuthTokenConfigurer;
http.apply(securityConfigurerAdapter);
}
Spring Security CSRF Documentation
You need to include the token in your requests. If you use Thymeleaf as your templating engine, this is handled automatically. The documentation also describes how to handle Ajax as well.
I've got spring security configured as
#Configuration
#EnableWebSecurity
#EnableGlobalMethodSecurity(securedEnabled = true, prePostEnabled = false)
public class WebSecurityConfig extends WebSecurityConfigurerAdapter {
#Override
protected void configure(HttpSecurity http) throws Exception {
http.csrf().disable()
.authorizeRequests()
.anyRequest().authenticated()
.and()
.jee()
.mappableRoles("ROLE1", "ROLE2");
}
}
And then #Secured annotations with roles on the rest endpoints.
Doesn't matter what I do I don't seem to be able to create a custom handler for authorization (i.e. a user logged in successfully but doesn't have the right role to access a particular endpoint) error events.
What I tried was:
An exception handler with #ExceptionHandler(value = AccessDeniedException.class) - doesn't get called. I understand that's by design, ok.
AuthenticationEntryPoint configured as
http.exceptionHandling().authenticationEntryPoint(new RestAuthenticationEntryPoint())
#Component( "restAuthenticationEntryPoint" )
public class RestAuthenticationEntryPoint implements AuthenticationEntryPoint {
#Override
public void commence( HttpServletRequest request, HttpServletResponse response,
AuthenticationException authException ) throws IOException {
// logging
}
}
-doesn't get called
ApplicationListener - I can see it's getting called on context closed, so it's registered correctly but not called on authorization error.
All I need is a simple handler to log unsuccessful authorization events.
It completely slipped my mind that the allowed roles are listed in web.xml as well for j2ee container authentication to work. So any user without a least one of those roles was just being rejected by the container.
Otherwise the first, simplest, method works fine. Hopefully my mistake will help someone in the future
Swagger works! I can interact with http://localhost:8090/sdoc.jsp and everything is fine.
I add the following to pom.xml...
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-security</artifactId>
</dependency>
I also add the following two files:
#Component
public class CustomAuthenticationProvider implements AuthenticationProvider {
#Override
public Authentication authenticate(Authentication authentication) throws AuthenticationException {
String name = authentication.getName();
String password = authentication.getCredentials().toString();
if( !Authenticate.authenticate(name, password) )
return null;
List<GrantedAuthority> grantedAuths = new ArrayList<>();
grantedAuths.add(new SimpleGrantedAuthority("ROLE_USER"));
Authentication auth = new UsernamePasswordAuthenticationToken(name, password, grantedAuths);
return auth;
}
#Override
public boolean supports(Class<?> authentication) {
return authentication.equals(UsernamePasswordAuthenticationToken.class);
}
}
and
#Configuration
#EnableWebMvcSecurity
public class WebSecurityConfig extends WebSecurityConfigurerAdapter {
#Override
protected void configure(HttpSecurity http) throws Exception {
http
.csrf()
.disable()
.authorizeRequests()
.anyRequest().permitAll()
.antMatchers("/**").authenticated().and()
.formLogin().loginPage("/login").permitAll().and()
.httpBasic()
;
}
#Autowired
public void configureGlobal(AuthenticationManagerBuilder auth) throws Exception {
auth.authenticationProvider(new CustomAuthenticationProvider());
}
}
At this point if I visit the same URL that was previously working I now instead get a response type of "text/plain" and instead of a pretty HTML looking browser I see source code.
If I revert the change and remove the two files from project and remove JAR file it works again.
How do I get Spring Security and Swagger to play nice? What am I doing wrong.
I suspect this is due to Spring-Security's effect on the content-type headers (http://docs.spring.io/spring-security/site/docs/3.2.0.CI-SNAPSHOT/reference/html/headers.html#headers-content-type-options).
From the docs -
Historically browsers, including Internet Explorer, would try to guess the content type of a request using content sniffing. This allowed browsers to improve the user experience by guessing the content type on resources that had not specified the content type. For example, if a browser encountered a JavaScript file that did not have the content type specified, it would be able to guess the content type and then execute it.
The problem with content sniffing is that this allowed malicious users to use polyglots (i.e. a file that is valid as multiple content types) to execute XSS attacks. For example, some sites may allow users to submit a valid postscript document to a website and view it. A malicious user might create a postscript document that is also a valid JavaScript file and execute a XSS attack with it.
Again, from the docs, in order to override the default -
#EnableWebSecurity
#Configuration
public class WebSecurityConfig extends
WebSecurityConfigurerAdapter {
#Override
protected void configure(HttpSecurity http) throws Exception {
http
// ...
.headers()
.contentTypeOptions();
}
}
Wow, I figured it was something along these lines. Thanks so much
When I tried this and it started working
.headers()
.disable()
I narrowed the default contentTypeOptions down to..
.headers()
//.contentTypeOptions() // If this is uncommented it fails.
.xssProtection()
.cacheControl()
.httpStrictTransportSecurity()
.frameOptions()
.and()