Different AuthenticationManager per path/route in spring security in MvC - rest

Preamble
Since there are a lot of questions on StackOverflow about this already, I first want to ensure that this is not a duplicate and differentiate.
This is about
Having 2(or more) different AuthenticationProviders in 2 different AuthenticationManagers to be used on different routes.
Using the methods in Spring Security 5.5 not 3.x
Using a non XML configuration based approach
So the question is not about:
How to include several AuthenticationProvideres in on AuthenticationManager to offer "alternative authentications" (which most questions tend to be)
Case
Assume one has 2 custom AuthenticationProviders: CATApiTokenProvider and DOGApiTokenProvider. It is by design that we not talk about AOuth/JWT/Basic/Form providers, since they offer shortcuts.
Now we have 2 REST API endpoints /dog/endpoint and /cat/endpoint.
Question
How would one properly implement this today, with Spring Security 5.5:
We want the authentication provider CATApiTokenProvider to only be able to authenticate requests on /cat/endpoint
We want the authentication provider DOGApiTokenProvider to only be able to authenticate requests on /dog/endpoint
So one cannot authenticate with a cat token on /dog/endpoint and neither with a dog token on /cat/endpoint.
My Ideas/Approaches
a) I understand that since I have custom Cat/Dog filters, one can use the AuthenticationManagerResolver and pass one instance into the filter when creating the bean. This resolver might look like
public AuthenticationManagerResolver<HttpServletRequest> resolver()
{
return request -> {
if (request.getPathInfo().startsWith("/dog/")) {
try {
return ???;
} catch (Exception exception) {
log.error(exception);
}
}
if (request.getPathInfo().startsWith("/cat/")) {
try {
return ???;
} catch (Exception exception) {
log.error(exception);
}
}
};
}
Two questions with that would be:
how to return different authentication managers here? How to instantiate 2 different AM with each one CatAP and DogAP? Currently I use public void configure(AuthenticationManagerBuilder auth) but as far as I understand, I would only configure 'the one' AuthenticationManager and I could add DogAP and CatAP there, but this would let as having 1 AM with 2 APs, so when using this AM i could auth with the dog token on the cat endpoint
is this really the right way to implement this? I would have expected to be able to provide the AM on the SecurityConfiguration level
b) Somehow instantiate 2 different AuthenticationManagers and then use the SecurityConfiguration to assign them to different matchers.
Two questions:
what is the right way to spawn 2 different AMs with different providers?
I cannot understand how I would add an AM for a spec
http.authorizeRequests()
.antMatchers("/dog/**")
.?

You can either publish multiple filter chains or wire your own AuthenticationFilter with an AuthenticationManagerResolver
You may use AuthenticationManagerResolver to return different AuthenticationManagers. Since Spring Security 5.4.0, we don't need to extend the WebSecurityConfigurerAdapter to configure our SecurityFilterChain anymore, you can instead define a bean of SecurityFilterChain type.
I'll go into detail on wiring your own AuthenticationFilter.
#Configuration
#EnableWebSecurity
public class SecurityConfig {
#Bean
public SecurityFilterChain apiSecurity(HttpSecurity http) throws Exception {
http.authorizeHttpRequests((authz) -> authz
.anyRequest().authenticated());
http.addFilterBefore(apiAuthenticationFilter(), UsernamePasswordAuthenticationFilter.class);
return http.build();
}
private AuthenticationFilter apiAuthenticationFilter() {
AuthenticationFilter authenticationFilter = new AuthenticationFilter(new ApiAuthenticationManagerResolver(), new BasicAuthenticationConverter());
authenticationFilter.setSuccessHandler((request, response, authentication) -> {});
return authenticationFilter;
}
public static class ApiAuthenticationManagerResolver implements AuthenticationManagerResolver<HttpServletRequest> {
private final Map<RequestMatcher, AuthenticationManager> managers = Map.of(
new AntPathRequestMatcher("/dog/**"), new DogAuthenticationProvider()::authenticate,
new AntPathRequestMatcher("/cat/**"), new CatAuthenticationProvider()::authenticate
);
#Override
public AuthenticationManager resolve(HttpServletRequest request) {
for (Map.Entry<RequestMatcher, AuthenticationManager> entry : managers.entrySet()) {
if (entry.getKey().matches(request)) {
return entry.getValue();
}
}
throw new IllegalArgumentException("Unable to resolve AuthenticationManager");
}
}
public static class DogAuthenticationProvider implements AuthenticationProvider {
#Override
public Authentication authenticate(Authentication authentication) throws AuthenticationException {
if (authentication.getName().endsWith("_dog")) {
return new UsernamePasswordAuthenticationToken(authentication.getName(), authentication.getCredentials(),
AuthorityUtils.createAuthorityList("ROLE_DOG"));
}
throw new BadCredentialsException("Username should end with _dog");
}
#Override
public boolean supports(Class<?> authentication) {
return true;
}
}
public static class CatAuthenticationProvider implements AuthenticationProvider {
#Override
public Authentication authenticate(Authentication authentication) throws AuthenticationException {
if (authentication.getName().endsWith("_cat")) {
return new UsernamePasswordAuthenticationToken(authentication.getName(), authentication.getCredentials(),
AuthorityUtils.createAuthorityList("ROLE_CAT"));
}
throw new BadCredentialsException("Username should end with _cat");
}
#Override
public boolean supports(Class<?> authentication) {
return true;
}
}
}
In the example above, we have two AuthenticationProviders, one for cat and other for dog. They are resolved upon an AntPathRequestMatcher matching for both /dog/** and /cat/** endpoints, inside the ApiAuthenticationManagerResolver. There is no need to defined an AuthenticationManager for each dog and cat, since AuthenticationProvider/Manager have the same interface.
The ApiAuthenticationManagerResolver is then wired inside an AuthenticationFilter in your filter chain.
You can also define two different filter chains for each endpoint, like so:
#Bean
public SecurityFilterChain dogApiSecurity(HttpSecurity http) throws Exception {
http.requestMatchers((matchers) -> matchers
.antMatchers("/dog/**"));
http.authorizeRequests((authz) -> authz
.anyRequest().authenticated());
http.httpBasic();
http.authenticationProvider(new DogAuthenticationProvider());
return http.build();
}
#Bean
public SecurityFilterChain catApiSecurity(HttpSecurity http) throws Exception {
http.requestMatchers((matchers) -> matchers
.antMatchers("/cat/**"));
http.authorizeRequests((authz) -> authz
.anyRequest().authenticated());
http.httpBasic();
http.authenticationProvider(new CatAuthenticationProvider());
return http.build();
}
Please, when defining multiple filter chains, the ordering is important, make use of the #Order annotation in those scenarios.
When you do http.requestMatcher(new AntPathRequestMatcher("/endpoint/**")); you are telling Spring Security to only call the filter chain when the request matches that path.
There is also a ticket within Spring Security's repository to provide a AuthenticationManagerResolver implementation which accepts Map<RequestMatcher, AuthenticationManager>, it would be nice if you think it makes sense, give a thumbs up there.

Related

Jersey client with RxJava conflicts with resteasy

I have a web application deployed to JBOSS. It contains dependency to jersey-rx-client-rxjava package and one of the packages has transient dependency to resteasy-jaxrs.
I have the following code.
RxObservable.newClient()
.target(fullURL)
.request()
.header("Authorization", "Bearer " + config.getApiKey())
.rx()
.post(javax.ws.rs.client.Entity.entity(context, MediaType.APPLICATION_JSON_TYPE), AIResponse.class)
.map(new Func1<AIResponse, String>() {
#Override
public String call(AIResponse res) {
return res.getType();
}
})
.subscribe(new Action1<String>() {
#Override
public void call(final String type) {
Log.info(type);
}
}, new Action1<Throwable>() {
#Override
public void call(final Throwable throwable) {
//async.resume(throwable);
Log.error(throwable.getMessage(), throwable);
}
}, new Action0() {
#Override
public void call() {
//async.resume(throwable);
Log.info("Done");
}
});
At this line, the following exception is thrown.
final JerseyInvocation invocation = (JerseyInvocation) getBuilder().build(name, entity);
Why does the build method return org.jboss.resteasy.client.jaxrs.internal.ClientInvocation, instead of JerseyInvocation?
2017-03-20 12:18:43,678 ERROR [com.optawork.bot.CustomResource] (default task-2) org.jboss.resteasy.client.jaxrs.internal.ClientInvocation cannot be cast to org.glassfish.jersey.client.JerseyInvocation: java.lang.ClassCastException: org.jboss.resteasy.client.jaxrs.internal.ClientInvocation cannot be cast to org.glassfish.jersey.client.JerseyInvocation
at org.glassfish.jersey.client.rx.rxjava.JerseyRxObservableInvoker$2.call(JerseyRxObservableInvoker.java:89)
at org.glassfish.jersey.client.rx.rxjava.JerseyRxObservableInvoker$2.call(JerseyRxObservableInvoker.java:83)
at rx.Observable.unsafeSubscribe(Observable.java:10142)
at rx.internal.operators.OnSubscribeMap.call(OnSubscribeMap.java:48)
at rx.internal.operators.OnSubscribeMap.call(OnSubscribeMap.java:33)
at rx.Observable.subscribe(Observable.java:10238)
at rx.Observable.subscribe(Observable.java:10205)
at rx.Observable.subscribe(Observable.java:10086)
at ai.api.AIDataService.converse(AIDataService.java:601)
Why does the build method return org.jboss.resteasy.client.jaxrs.internal.ClientInvocation, instead of JerseyInvocation
It's just how the JAX-RS Client API is designed. When we try to call ClientBuilder.newBuilder (which is done internally), the JAX-RS API does a service lookup for any implementation of the JAX-RS Client API. If there is none, it falls back to Jersey. The problem is that when the service lookup is done, RESTEasy's client is found on the classpath.
The Jersey RX API has a from(Client) method that we can user, instead of the default newClient. This will allow us to pass an explicit JerseyClient instead of using the JAX-RS API ClientBuilder.newBuilder/newClient
// actual JerseyClient which implements Client
Client client = new JerseyClientBuilder().build();
RxObservable.from(client)
JerseyClientBuilder has pretty much the same API as the JAX-RS ClientBuilder, so you can use it pretty much the same way.

Creating Custom Operators in RxJava2?

I'm having trouble finding an example of how to make a custom operator with RxJava 2. I've considered a few approaches:
Using Observable.create, and then flatMaping on it from the source observable. I can get this working, but it doesn't quite feel right. I end up creating a static function which I provide the source Observable, and then flatMap on the source. In the OnSubscribe, I then instantiate an object that I pass the emitter to, which handles and manages the Observable / Emitter (as it's not trivial, and I want everything as encapsulated as possible).
Creating an ObservableOperator and providing it to Observable.lift. I can't find any examples of this for RxJava 2. I had to debug into my own example to make sure my understanding of upstream and downstream were correct. Because I can't find any examples or documentation on this for RxJava 2 I'm a little worried I might accidentally do something I'm not supposed to.
Create my own Observable type. This seems to be how the underlying operators work, many of which extend AbstractObservableWithUpstream. There is a lot going on here though, and it seems easy to miss something or do something I shouldn't. I'm not sure if I'm supposed to take an approach like this or not. I stepped myself through the mental process, and it seems like it can get hairy pretty quickly.
I'm going to proceed forward with option #2, but thought it worthwhile to ask what the supported method for doing this was in RxJava2 and also find out if there was any documentation or examples for this.
Writing operators is not recommended for beginners and many desired flow patterns can be achieved via existing operators.
Have you looked at RxJava's wiki about writing operators for 2.x? I suggest reading it from top to bottom.
using create() is possible but most people use it to emit the elements of a List with a for-each loop, not recognizing that Flowable.fromIterable does that.
We kept this extension point although RxJava 2 operators don't use lift() themselves. If you want to avoid some boilerplate with option 3. then you may try this route.
This is how RxJava 2 operators are implemented. AbstractObservableWithUpstream is a small convenience and not necessary for external implementors.
This may help you. I implement operator RxJava2 to handle APiError. I used lift operator.
See the example.
public final class ApiClient implements ApiClientInterface {
...
#NonNull
#Override
public Observable<ActivateResponse> activate(String email, EmailData emailLinkData) {
return myApiService.activate(email, emailData)
.lift(getApiErrorTransformer())
.subscribeOn(Schedulers.io());
}
private <T>ApiErrorOperator<T> getApiErrorTransformer() {
return new ApiErrorOperator<>(gson, networkService);
}
}
And then you can find custom operator
public final class ApiErrorOperator<T> implements ObservableOperator<T, T> {
private static final String TAG = "ApiErrorOperator";
private final Gson gson;
private final NetworkService networkService;
public ApiErrorOperator(#NonNull Gson gson, #NonNull NetworkService networkService) {
this.gson = gson;
this.networkService = networkService;
}
#Override
public Observer<? super T> apply(Observer<? super T> observer) throws Exception {
return new Observer<T>() {
#Override
public void onSubscribe(Disposable d) {
observer.onSubscribe(d);
}
#Override
public void onNext(T value) {
observer.onNext(value);
}
#Override
public void onError(Throwable e) {
Log.e(TAG, "onError", e);
if (e instanceof HttpException) {
try {
HttpException error = (HttpException) e;
Response response = error.response();
String errorBody = response.errorBody().string();
ErrorResponse errorResponse = gson.fromJson(errorBody.trim(), ErrorResponse.class);
ApiException exception = new ApiException(errorResponse, response);
observer.onError(exception);
} catch (IOException exception) {
observer.onError(exception);
}
} else if (!networkService.isNetworkAvailable()) {
observer.onError(new NetworkException(ErrorResponse.builder()
.setErrorCode("")
.setDescription("No Network Connection Error")
.build()));
} else {
observer.onError(e);
}
}
#Override
public void onComplete() {
observer.onComplete();
}
};
}
}

Capture exception thrown by Hystrix fallback?

I'm designing a service facade and I have a method signature that looks like this:
public Policy getPolicy(long policyId) throws PolicyNotFoundException
If nothing bad happens then a Policy object (simple POJO) is returned. If the requested policy is not found then a checked exception PolicyNotFoundException is thrown (just as a reference - we follow this article when it comes to best practices on exception handling within an application).
The layer above the service facade layer (in this case a Spring MVC RestController) knows how to handle such a PolicyNotFoundException and return an appropriate payload.
I'm trying to incorporate this into a HystrixCommand by doing something like this:
#HystrixCommand(groupKey = "PolicyService", fallbackMethod = "getPolicySafe", ignoreExceptions = { PolicyNotFoundException.class })
public Policy getPolicy(long policyId) throws PolicyNotFoundException {
LOGGER.info("Getting policy {}", policyId);
// Simulate some error condition for testing purposes
throw new RuntimeException("Something happened!");
}
private Policy getPolicySafe(long policyId, Throwable t) throws PolicyNotFoundException {
LOGGER.warn("Falling back to circuit-breaker for getting policy {}", policyId, t);
throw new PolicyNotFoundException(policyId);
}
Basically I want my circuit breaker to simply behave as if the policy wasn't found by the original lookup. The problem I'm having with this though is the exception I throw from the fallback method is getting lost in translation somewhere. The exception I end up seeing in the layer above is the RuntimeException thrown by the command method and not the exception thrown by the fallback method. Is there a way around this? I don't want to change the contract of my original method either nor do I want the layer above this to know anything other than to have to catch PolicyNotFoundException in the case a policy isn't found. Whatever is needed here should be captured within this service facade layer.
Any and all help would be greatly appreciated. Thanks!
So based on the link #spencergibb gave - I may have found a solution after upgrading to Hystrix 1.5.7. This code works as expected
PolicyRestController.java
#RestController
#RequestMapping("/policies")
public class PoliciesApi {
private static final Logger LOGGER = LoggerFactory.getLogger(PoliciesApi.class);
#Autowired
private PolicyService policyService;
#RequestMapping(value = "/{policyId}", method = RequestMethod.GET, produces = { MediaTypes.POLICY_JSON_VALUE, MediaTypes.POLICY_XML_VALUE })
public Policy getPolicy(#PathVariable long policyId) {
try {
// This just shown for simplicity. There is more to this method (input validation/etc)
return this.policyService.getPolicy(policyId);
}
catch (PolicyNotFoundException ex) {
// NotFoundException is a RuntimeException annotated with #ResponseStatus(HttpStatus.NOT_FOUND)
// So the service returns a 404 to the client
LOGGER.info("Policy {} wasn't found", ex.getPolicyId(), ex);
throw new NotFoundException(String.format("Policy %s was not found", ex.getPolicyId()));
}
}
}
PolicyService.java
public interface PolicyService {
#Cacheable("allPolicies")
public List<Policy> getPolicies();
#Cacheable("policies")
public Policy getPolicy(long policyId) throws PolicyNotFoundException;
}
PolicyServiceImpl.java:
#Service
public class PolicyServiceImpl implements PolicyService {
#HystrixCommand(groupKey = "PolicyService", fallbackMethod = "getPolicySafe", ignoreExceptions = { PolicyNotFoundException.class })
public Policy getPolicy(long policyId) throws PolicyNotFoundException {
LOGGER.info("Getting policy {}", policyId);
// Simulate some error condition for testing purposes
throw new RuntimeException("Something happened!");
}
#HystrixCommand(groupKey = "PolicyService", ignoreExceptions = { PolicyNotFoundException.class }, raiseHystrixExceptions = { HystrixException.RUNTIME_EXCEPTION })
private Policy getPolicySafe(long policyId) throws PolicyNotFoundException {
// Here is we hit our fallback we want to log a warning & simply act as if the policy wasn't found by throwing the same contingency exception as the API does
LOGGER.warn("Falling back to circuit-breaker for getting policy {}", policyId);
throw new PolicyNotFoundException(policyId);
}
}
While your solution might work for you I've noticed some weirdness in your code (I can't check my assumptions so I would like to ask you to check this).
Try to avoid using checked exceptions in your code because it's
awkward to maintain.
Based on your code you will never catch
"PolicyNotFoundException" since you're using raiseHystrixExceptions = { HystrixException.RUNTIME_EXCEPTION } which means that you won't to get your custom exception so that HystrixRuntimeException will be propagated. Try to rewrite your code as follows so it should
simplify the code and maybe fix some of your problems:
#Service
public class PolicyServiceImpl implements PolicyService {
#HystrixCommand(groupKey = "PolicyService", fallbackMethod = "getPolicySafe")
public Policy getPolicy(long policyId) throws PolicyNotFoundException {
LOGGER.info("Getting policy {}", policyId);
throw new PolicyNotFoundException(); // throw real PolicyNotFoundException if policy is absent for the given id
}
#HystrixCommand(groupKey = "PolicyService")
private Policy getPolicySafe(long policyId) throws PolicyNotFoundException {
// Here is we hit our fallback we want to log a warning & simply act as if the policy wasn't found by throwing the same contingency exception as the API does
LOGGER.warn("Falling back to circuit-breaker for getting policy {}", policyId);
throw new PolicyNotFoundException(policyId);
}
}
This is the default behavior of hystrix. "If command has a fallback then only first exception that trigers fallback logic will be propagated to caller"
See the error propagation section here.
I do this:
#Component
public class HystrixClient {
#HystrixCommand(ignoreExceptions = {ClientArgumentException.class})
public POJO getPojo(String id)
throws ClientNoDataFoundException, ClientArgumentException, ClientGeneralException {
//call my service and return POJO
}
}
#Component
public TrueClientUsedForAnotherSerivce {
#Autowired
HystrixClient hystrixClient;
public POJO getPojo(String id)
throws ClientNoDataFoundException, ClientArgumentException, ClientGeneralException, ClientOpenCircuitException {
try {
POJO result = hystrixClient.getCellular(id);
return result;
}
catch(HystrixRuntimeException e) {
LOG.debug("The circuit is open");
throw new ClientOpenCircuitException("Open circuit");
}
}
It only works if #HystrixCommand method is in another class.

Lifestyle of Castle.Windsor interceptors

In my [Webmethod]s I have code like this.
var something = container.ResolveSomething();
something.Run();
All registered components except one have lifestyle defined as PerWebRequest. One is registered as Singleton (logger).
For some of components I have defined and configured Interceptor that will log method calls and their results.
My question is: Will I have problems if I register this Interceptor with Lifestyle PerWebRequest? Documentation advices to make all Interceptors Transient and use other lifestyles if we are really sure we want to do it. If I register Interceptors with lifestyle Transient any of my about 100 methods will have to look like this.
IComponent component = null;
try
{
component = container.ResolveComponent();
compoment.Run();
}
finally
{
container.Release(component);
}
So more boilerplate then real code.
Here is my interceptor:
public class LoggingInterceptor : IInterceptor
{
private readonly ILogger logger;
public LoggingInterceptor(ILogger logger)
{
this.logger = logger;
}
public void Intercept(IInvocation invocation)
{
var call = string.Format("{0}.{1}({2})", invocation.TargetType.FullName, invocation.Method.Name, string.Join(", ", invocation.Arguments.Select(arg => arg.ToString()).ToArray()));
try
{
logger.Info(call);
invocation.Proceed();
logger.Info("Result: " + call + " = " + invocation.ReturnValue);
}
catch (Exception e)
{
logger.Error(call, e);
throw;
}
}
}
I know WCF is better prepaired for IoC but I have to stay with ASP.NET WebServices.
My understanding is that even if the Interceptor is transient, it's lifetime is bound with intercepted component. It will be released with intercepted component because it is tracked by container.

gwt - servlet path + url

I need to read data from an xml file that is under the WAR directory.
I'm using RequestBuilder for creating the GET request.
It looks like this:
RequestBuilder requestBuilder = new RequestBuilder(RequestBuilder.GET,"customerRecord.xml");
try {
requestBuilder.sendRequest(null, new RequestCallback() {
public void onError(Request request, Throwable exception) {
requestFailed(exception);
}
public void onResponseReceived(Request request,Response response) {
renderXML(response.getText());
}
});
} catch (RequestException ex) {
requestFailed(ex);
}
Now, the thing is that I don't want to load all of the data. I want to send a parameter that tells the server which part to bring, (let's say - how many lines of data) and then override the doGet method of the servlet and deal with the parameter.
I have 2 questions:
1) how do I declare the path of the servlet? where is the connection between the servlet and the request??
2) What do I write in the url of the RequestBuilder (instead of "customerRecord.xml")? do I need to refer to the servlet there or I can keep it like
May be You mean GWT Service?
You need to create 2 interfaces - Service and ServiceAsync and implementation of Service in server package (on same level as client package). Then You define implementation as servlet (in my JBoss 7.1 it just annotation. in older version servlet mapping):
#WebServlet(name="YourService", urlPatterns={"/%module%/YourService"})
public class YourServiceImpl extends RemoteServiceServlet implements YourService
in Your modeule.xml write:
<servlet path="/YourService" class="org.name.YourServiceImpl"/>
and in the end You can call this service from Your code
YourService.App.getInstance().getSomething(new AsyncCallback<Collection<Something>>() {
#Override
public void onFailure(Throwable caught) {
new MessagePopup("Error: " + caught.getMessage()).center();
}
#Override
public void onSuccess(Collection<Something> result) {
}
});
Interfaces You can create from Your beloved IDE. It's much simpler)
One think which still bothering me - I cannot specify path for servlet in another module.