I'm using google-api-services-admin-directory in a project, and the docs recommend using GoogleNetHttpTransport.newTrustedTransport() as the HTTP transport, but I kept getting certificate errors that I think were cause by my proxy. I switched over to new NetHttpTransport() instead, and everything now works fine, but I just want to make sure I'm not compromising on security by doing so.
Any thoughts?
Thanks,
Rhys
Hope you don't mind the late reply. I was having the same issue and did some research.
According to the Java Docs of Google API the GoogleNetHttpTransport.newTrustedTransport() takes into account website certificates to use for Google APIs. However:
This helper method doesn't provide for customization of the NetHttpTransport, such as the ability to specify a proxy.
This confirms what you were saying about the certificates errors you are getting being caused by your proxy.
It is recommended, security wise, that instead of new NetHttpTransport() to use the NetHttpTransport.Builder in order to get a Trusted Transport for a specified proxy, as show on the link's example:
static HttpTransport newProxyTransport() throws GeneralSecurityException, IOException {
NetHttpTransport.Builder builder = new NetHttpTransport.Builder();
builder.trustCertificates(GoogleUtils.getCertificateTrustStore());
builder.setProxy(new Proxy(Proxy.Type.HTTP, new InetSocketAddress("127.0.0.1", 3128)));
return builder.build();
}
Related
A small question regarding Netty and io.netty.handler.ssl.SslContext
In Tomcat and org.apache.http.ssl.SSLContexts, we have the possibility to perform the following:
HttpClient httpClient = HttpClients.custom()
.setSSLContext(SSLContexts.custom()
.loadKeyMaterial(someKeystorePropertlyInitialized)
.loadTrustMaterial(someTruststorePropertlyInitialized)
.build())
.setSSLHostnameVerifier(NoopHostnameVerifier.INSTANCE)
.build();
(Appreciate if we can leave the fonts and not wrap inside a code block)
This can for instance fix issues such as Caused by: java.security.cert.CertificateException: No subject alternative DNS name matching xxx found
(This question is not about if NoopHostnameVerifier.INSTANCE is the proper way to fix this.)
My question is, what is the equivalent in Netty of .setSSLHostnameVerifier(NoopHostnameVerifier.INSTANCE), without .trustManager(InsecureTrustManagerFactory.INSTANCE), because I have a real trust store, I just want to skip the host name, not everything
Maybe something with reactor.netty.http.client.HttpClient; HttpClient.create() ?
Actually, Netty has hostname verification turned off by default -- see this issue. It looks like the library you're using (reactor-netty) might have it turned on. There appears to be a similar issue on reactor-netty's github which points to the solution, but the code snippet provided seems to do more than what's necessary. Essentially, all you need is to access the SSLEngine from the SslHandler and make sure the endpoint identification algorithm is empty/null:
HttpClient.create().secure(
ssl -> ssl.sslContext(sslContext)
.handlerConfigurator(handler-> {
SSLEngine engine = handler.engine();
SSLParameters params = new SSLParameters();
// ... set other SSL params
params.setEndpointIdentificationAlgorithm(null);
})
);
My Guidelines
If followed this Google documentation about verifying Google-Account-Tokens on the server side, but I am kinda confused.
My Problem
GoogleIdTokenVerifier googleIdTokenVerifier = new GoogleIdTokenVerifier.Builder(new NetHttpTransport(), new JacksonFactory())
.setAudience(Collections.singletonList(CLIENT_ID))
.build();
In this piece of code I figured out that the transport and jsonFactory arguments can be filled as new NetHttpTransport() and new JacksonFactory() here. It also describes how to get AudienceString, but I couldn't figure out what it is for. I couldn't test it, but my question is if I can use it without .setAudience() or if I need it and what it is for.
In .setAudience() you have to pass all client ID's. You can get the ID for your client from the Credentials Page. It's explained here.
Thanks to #StevenSoneff.
If you didn't get the basic concept
For every client you want your server to accept, you need to create a project in the `Developer Console`. Clients are differentiated by their `SHA-1` fingerprint. You can for example have a debug project (will take your debug fingerprint) and a release one. To make both work, you have to add both `ID`'s to your server's `GoogleIdTokenVerifier`'s `.setAudience()`.
In my case, If you're using Firebase to get the id token on Android or iOS. You should follow these instructions to verify it on your backend server.
Verify ID tokens using a third-party JWT library
For me, I'm using Google OAuth Client as the third-party library so it's easy to use.
But it's a little bit different from this document.
Verify the Google ID token on your server side
The CLIENT_ID is your firebase project ID.
The Issuer has to be set as https://securetoken.google.com/<projectId>.
You need to use GooglePublicKeysManager and call setPublicCertsEncodedUrl to set it as https://www.googleapis.com/robot/v1/metadata/x509/securetoken#system.gserviceaccount.com
GooglePublicKeysManager manager = new GooglePublicKeysManager.Builder(HTTP_TRANSPORT, JSON_FACTORY)
.setPublicCertsEncodedUrl(PUBLIC_KEY_URL)
.build();
GoogleIdTokenVerifier verifier = new GoogleIdTokenVerifier.Builder(manager)
.setAudience(Collections.singletonList(FIREBASE_PROJECT_ID))
.setIssuer(ISSUER)
.build();
If you have multiple issuers, then you have to create GoogleIdTokenVerifier for each one.
I have the following code:
Class<OAuthHandler> _tempClass = (Class<OAuthHandler>) Class.forName(providerClass);
Constructor<OAuthHandler> oAuthHandlerConstructor = _tempClass.getDeclaredConstructor(Vertx.class);
OAuthHandler oAuthHandler = oAuthHandlerConstructor.newInstance(vertx);
OAuth2Auth oAuth2Auth = oAuthHandler.getoAuth2Auth();
/* AccessToken accessToken = */ oAuth2Auth.getToken(oAuthHandler.getTokenConfig(code), accessTokenResponse -> {
if (accessTokenResponse.failed()) {
System.out.println("Failed to obtain token");
} else {
AccessToken accessToken = accessTokenResponse.result();
// Return the token? Somehow.
}
});
The oAuthHandler is a provider specific implementation providing some config etc based on the provider name, and simply wraps around the methods provided by the vertx-auth-oauth2 library.
I would like to use the access token after it returns, but not inside the getToken Lambda, so I can get any information about the user I need.
Ive seen some things about Vert.x Futures, but unsure if I can even use them here, any suggestions or examples to solutions would be much appreciated!
Vert.x OAuth2 support is not vendor specific. According to the documentation http://vertx.io/docs/vertx-web/java/#_oauth2authhandler_handler you will see that the same code can handle the following providers:
Google
Twitter
Github
LinkedIn
Facebook
Keycloak
and as soon as 3.4 is released some small fixes will make it also compatible with:
Azure AD
Also the handler is generic so if you have your own provider you can also use it.
Now regarding the second part of the question how to use the the token directly, then you probably do not want the OAuth2Handler since it hides all this from you and you want to interact with OAuth2 client directly:
http://vertx.io/docs/vertx-auth-oauth2/java/#_getting_started
From the documentation above there is an getting started code example that allows you to interact with the get token without using reflection. This is way better since you will not be hurt if internal api's change.
Applications such a Google's Chrome and IE can transparently handle Kerberos authentication; however I can not find a "simple" Java solution to match this transparency. All of the solutions I have found require the presence of a krb5.conf file and a login.conf file which nether of the above apps seem to require.
What is the best way to build a Java application with Kerberos SSO capabilities that just work?
[update]: to be clear I need a CLIENT side solution for creating tickets not validating them. Also, it seems that SPNEGO is the default "wrapper" protocol that will eventually delegate to Kerberos but I need to be able to handle the SPNEGO protocol as well.
There is now a simple solution for this using the Apache HTTP Components Client 4.5 or greater. This is still marked as experimental in 4.5 so your milage may vary, but this is working fine for me in an enterprise context.
In addition to the HC 4.5 client jars you will need to have the httpclient-win, jna and jna-platform jars on your classpath, as provided with http-component-client. You then construct a Kerberos enabled HC-client as follows:
CloseableHttpClient httpclient = WinHttpClients.createDefault();
Or using the builder:
HttpClientBuilder clientBuilder = WinHttpClients.custom();
Which can then be customised as required before building the client:
CloseableHttpClient client = clientBuilder.build();
This solution works without any external configuration, and most importantly solves the issue where the in-built JRE mechanism breaks for users with local Admin rights on Windows 7+. This is possible because the Kerberos ticket is being retrieved directly from the SSPI API via JNA, rather than going through the GSSAPI provided by the JRE.
Example code from the http-components team
This was all made possible by the good work of Daniel Doubrovkine Timothy Wall
and Ryan McKinley
Adding to David Roussels answer on url specific http based kerberos authentication:-
The reason why your code works is because your target SPN(server side principal) is configured to with HTTP/serverhostname.realm.com#DOMAIN.COM. In that case it will work because you are not explicitly setting the token. URLConnection internally sets a token with that SPN
1 Perform steps(from my previous answer) to get a subject
2 Use gss api init sec context to generate a context token. There are numerous tutorials out there for this step
3 Base 64 encode the token
4 Attach the token to urlconnection:-
URL url = new URL("http://myhost/myapp")
HttpURLConnection urlConn = (HttpURLConnection)url.openConnection(); =
urlConn.setRequestProperty("Authorization", "Negotiate " + encodedToken);
5 Implement a priviledged action:-
//this internally calls the getInputStream
public class PrivilegedGetInputStream implements PrivilegedExceptionAction<InputStream>
6 Wrap the whole thing in Subject.doAs
//use prev answer instructions to get subject
Subject.doAs(subject, new PrivilegedGetInputStream(urlConnection)
Oracle has an example using Java's SaslClient. I'm not a Java programmer, but when I pointed this out once to someone who is, they were able to make it work pretty quickly. It may still require a "conf" file somewhere (n.b. Kerberos uses environment variables, often starting with KRB5_, to know where to look for such files). Also note that Kerberos itself does not include a transport of any kind--your app needs to know how to send and receive the Kerberos payloads the way the server expects (and this is different depending on the server you are trying to authenticate with).
Edit: you edited your question, so here's a link related to SPNEGO in Java which might be of some use:
http://download.oracle.com/javase/6/docs/technotes/guides/security/jgss/lab/part5.html
You don't actually need to do anything. In Java 6, on a Windows client machine you can do this:
new URL("http://myhost/myapp").openStream();
And negotiate authentication just works. At least it does for me. And the server I tested on only supports Negotiate, not NTLM auth.
Ok if you want to avoid using a login.conf file you need to code differently:-
//define your own configuration
import javax.security.auth.login.Configuration;
public class CustomLoginConfiguration extends Configuration
//pass certain parameters to its constructor
//define an config entry
import javax.security.auth.login.AppConfigurationEntry;
private AppConfigurationEntry configEntry;
//define a map of params you wish to pass and fill them up
//the map contains entries similar to one you have in login.conf
Map<String, String> params = new HashMap<String, String>();
//define the configuration
configEntry = new AppConfigurationEntry(
"com.sun.security.auth.module.Krb5LoginModule",
AppConfigurationEntry.LoginModuleControlFlag.REQUIRED, params);
//implement getappconfig method
public AppConfigurationEntry[] getAppConfigurationEntry() {
return new AppConfigurationEntry[] { configEntry };
}
Now once you are done with this definition you can use this in you use this to fetch tickets from kdc
//get ticket in login context
LoginContext lc = null;
lc = new LoginContext("lc", null, callback, new CustomLoginConfiguration(argumentlist));
lc.login();
Now from here on you can fetch jaas subject and can basically do a ton of authentication stuff.
In case you need further pointers just leave a comment.
You can use system properties instead of config files to specify the KDC hostname and service name, but those things (at least) are mandatory....
Waffle will actually give you the information you need to set most of the properties, even if it won't get you a ticket. Look at the WindowsAuthProviderImpl class (the Waffle.chm help file shows the API).
I use JAAS do obtain a service ticket from Active Directory in two steps:
Use Krb5LoginModule to retrieve the cached TGT and add it to the Subject.
Use the Subject and GSS-API to retrieve a service ticket from the KDC.
There's a lot of good information and example code at The Java Way of Active Directory.
I created a small tool to simplify connecting with httpclient to kerberos, you might want to give it a try.
https://github.com/DovAmir/httpclientAuthHelper
DefaultHttpClient httpclient = new DefaultHttpClient();
AuthUtils.securityLogging(SecurityLogType.KERBEROS,true);
CredentialsUtils.setKerberosCredentials(client, new UsernamePasswordCredentials("xxx", "xxx"), "domain", "kdc");
client.executeMethod(httpget);
Use WAFFLE
Here's a good blog post on having a java client to use with Kerberos
http://sachithdhanushka.blogspot.com/2014/02/kerberos-java-client-configuration.html
I'm building an Eclipse plugin that talks to a REST interface which uses Basic Authentication. When the authentication fails I would like to popup my plugin's settings dialog and retry. Normally I could use the static Authenticator.setDefault() to setup an authenticator for all HttpURLConnection's for this, but since I am writing a plugin I don't want to overwrite Eclipse's default Authenticator (org.eclipse.ui.internal.net.auth);
I thought of setting my custom Authenticator before loading and putting Eclipse's default back afterwards, but I imagine this will cause all sorts of race issues with multithreading so I quickly lost that notion.
Google searches yield all sorts of results basically telling me it's not possible:
The Java URLConnection API should have a setAuthenticator(Authenticator) method for making it easier to use this class in multi-threaded context where authentication is required.
Source
If applications contains few third party plugins and each plugin use its own Authenticator what we should do? Each invocation of "Authenticator.setDefault()" method rewrite previously defined Authenticator...
Source
Are there any different approaches that might help me overcome this issue?
If it is not possible with HttpURLConnection I would suggest using the httpclient library from Apache.
A quick example:
HttpClient client = new HttpClient();
client.getState().setCredentials(AuthScope.ANY, new UsernamePasswordCredentials("test","test"));
GetMethod getMethod = new GetMethod("http://www.example.com/mylogin");
client.executeMethod(getMethod);
System.out.println(getMethod.getResponseBodyAsString());
Another approach would be to perform the basic authentication yourself on the connection.
final byte[] encodedBytes = Base64.encodeData((username + ':' + new String(password)).getBytes("iso-8859-1"));
final String encoded = new String(encodedBytes, "iso-8859-1");
connection.setRequestProperty("Authorization", "Basic " + encoded);
This would also have the advantage of not requiring an unauthenticated request to receive a 401 before providing the credential on a subsequent request. Similar behavior can be leveraged in the apache http-client by requesting preemptive authentication.