I have an e2e test for a rather large application. It offers multiple ways of authenticating an user, one of which is using a google account by aquiring a OAuth2 token. Today it runs naively by manipulating the logon page that pops up and enters the logon details.
As expected, at some point google decided to throw this in my face:
How can I skip the manual logon procedure, grab an Oauth token and procede to my redirect url, and verify that I indeed arrive at the correct place in my application with selenium?
Offline authentication seems to be what I am looking for, however webdriver cannot do POST requests. I am able to acquire the token manually, but there's no way to pass it on to webdriver in any meaningful way as it still can't do anything with it.
If possible I would prefer to not touch the application codebase.
I am using webdriverio as my webdriver implementation, running as a nodejs task. Interestingly the robot filter appears rarely on my local machine, but seems to always trigger on the circleci instance it's run on.
I'm simply using my app endpoint to get a token from google, then using the JavaScript executor I set the token in the local storage.
Something like this.
driver.manage().deleteAllCookies();
JavascriptExecutor executor = (JavascriptExecutor) driver;
executor.executeScript("window.localStorage.setItem('token', ' " + apiRequest.getToken() + "');");
I hope this is what are you looking for.
Related
I'm working on some app which should use Google Maps JS API (browser version).
And there exists one problem I've got very concerned about.
API keys have restriction to be used only from your domain, however, any request from your domain (for example, from code inspector) is considered a valid request. So, anyone can make a simple script and kick my quota out easily.
So, here is my question:
Is there any option or command to run to block such activity ? Like the script will load just one instance and then will not accept creation of a new one or something like that.
P.S. I know about free quota for mobile versions of API, but I need the browser to work too. Obviously, I don't load this in any public area, but anyone can pretend to be a client and even order some service for couple bucks, but then run the script to make an impact for thousands ((
Scenario: Auth0 Single Page application client. .NET Web API and Angular SPA both configured to use this client. Works great.
I'd like to add Azure API Management as a layer in front of the API. Have set up the API in the Management Portal, updated SPA to call API, tested calls from SPA, works great.
Now, I'd like to configure API Management Portal with the right security settings such that people can invoke API calls from the Developer Portal. I've used this [https://auth0.com/docs/integrations/azure-api-management/configure-azure] as a guide.
Where I'm at:
From the Developer portal, I can choose Authorization Code as an Auth type, go through a successful sign-in process with Auth0 and get back a Bearer token. However, calls made to the API always return 401. I think this is because I'm confused about how to set it up right. As I understand it:
either I follow the instructions and setup a new API client in Auth0, but if that's the case then surely it's not going to work, because tokens generated from one client aren't going to work against my SPA client? (or is there something I need to change to make it work)
or, how should I configure Azure API Management to work with a SPA application. (this would be my preferred method, having two clients in Auth0 seems 'messy'). But, don't I need an 'audience' value in my authorization endpoint URL? How do I get that?
If anyone has done this, would very much appreciate some guidance here.
Well, I didn't think I'd be back to answer my own question quite so soon. The reason is mostly rooted in my general ignorance of this stuff, combined with trying to take examples and fuse them together for my needs. Posting this to help out anyone else who finds themselves here.
Rather than take the Single Application Client in Auth0 and make it work with Azure API Management, I decided to go the other way, and make the non-interactive Client work with my SPA. This eventually 'felt' more right: the API is what I'm securing, and I should get the API Management portal working, then change my SPA to work with it.
Once I remembered/realised that I needed to update my audience in the API to match the audience set in the Client in Auth0, then the Management Portal started working. Getting the SPA to work with the API then became a challenge: I was trying to find out how to change the auth0 angular code to pass an audience to match the one the API was sending, but it kept sending the ClientID instead. (by the way, finding all that out was made easier by using https://jwt.io/ to decrypt the Bearer tokens and work out what was happening - look at the 'aud' value for the audience.
In the end, I changed my API, in the new JwtBearerAuthenticationOptions object, the TokenValidationParameters object (of type TokenValidationParameters) has a property ValidAudiences (yes, there is also a ValidAudience property, confusing) which can take multiple audiences. So, I added my ClientID to that.
The only other thing I then changed (which might be specific to me, not sure) is that I had to change the JsonWebToken Signature Algorithm value in Auth0 for my non-interactive client (advanced settings, oAuth tab) from HS256 to RS256.
With all that done, now requests from both the API Management Portal, and my SPA work.
Curious to know if this is the "right" way of doing it, or if I've done anything considered dangerous here.
Since you're able to make the validation of the jwts with the .Net API work, Only few changes are actually necessary to get this working with Azure API Management.
In API management,
Create a validate-jwt inbound policy on an Operation (or all operations)
set the audiences and issuers the same as what you've used with your .NET web api. (you can check the values in Auth0 portal if you don't know this yet)
The important field that is missing at this point is the Open ID URLs since auth0 uses RS256 by default. The url can be found in you Auth0 portal at: Applications -> your single page application -> settings -> Scroll down, Show Advanced Settings -> End points. Then copy the OpenID Configuration
Here's the reference for API management's requirement for JWT tokens
optional reading
Background: This is my first standalone web development project, and my only experience in Meteor is building the Discover Meteor app over the last summer. I come from about a year of CS experience as a side interest in school, and I am most comfortable with C and C++. I have experience in python and java.
Project so far: I'm creating a calendar management system (for fun). Using accounts-google, I have created user accounts that are authenticated through google. I have requested the necessary permissions that I need for my app, including 'identity' and 'calendar read/write access'. I've spent the last week or so trying to get over this next hurdle, which is actually getting data from google.
Goal: I'd like to be able to make an API call to Calendar.list using a GET request. I've already called meteor add http to add the GET request functionality, my issue comes with the actual implementation.
Problem: I have registered my app on the developer console and set up Accounts using the client ID and secret, but I have not been able to find/generate my 'API key' for use in the request. Here is the google guide for creating the access token by using my (already) downloaded private key. I'm having a hard time wrapping my head around an implementation on the server side using JS because I don't have a lot of experience with what is mentioned in the HTTP/REST portion of the implementation examples. I would appreciate some help on how to implement a handshake and receive an access token for use in my app. If there is a call I can make or some package that will handle the token generation for me, that would be even better than implementation help. I believe an answer to this would also benefit this other question
The SO answer that I've been referring to so far: https://stackoverflow.com/a/14543159/4259653 Some of it is in spanish but it's pretty understandable code. He has an API key for his request, which I asked this question to help me with. The accounts-google documentation isn't really enough to explain this all to me.
Also an unrelated small question: What is the easiest way to deal with 'time' parameters in requests. I'm assuming JS has some sort of built-in functionality that I'm just not aware of yet.
Thanks for your research. I have also asked a very similar question, and right now I am looking into the package you recommend. I have considered this meteor-google-api package, but it looks abandoned.
Regarding your question about time manipulation, I recommend MomentJS. There are many packages out there; I am using meteor add mrt:moment
EDIT: MomentJS now has an official package for Meteor, so use meteor add momentjs:moment instead of the mrt command above
Below is a snippet of what moment can do. More documentation here.
var startTimeUTC = moment.utc(event.startTime, "YYYY-MM-DD HH:mm:ss").format();
//Changes above formatting to "2014-09-08T08:02:17-05:00" (ISO 8601)
//which is acceptable time format for Google API
So I started trying to implement all of this myself on the server side, but was wary of a lot of the hard-coding I was doing and assumptions I was making to fill gaps. My security prof. used to say "never implement encryption yourself", so I decided to take another gander for a helpful package. Revising search criteria to "JWT", I found jagi's meteor-google-oauth-jwt on Atmosphere. The readme is comprehensive and provides everything I need. Following the process used in The Google OAuth Guide, an authorization request can be made and a key generated for making an API call.
Link to Atmosphere: https://atmospherejs.com/jagi/google-oauth-jwt
Link to Repo: https://github.com/jagi/meteor-google-oauth-jwt/
I will update this answer with any additional roadblocks I hit in the Google API process and how I solved them:
Recently, I've been running into problems with the API request result. I get an empty calendarlist back from the API call. I suspect this is becuase I make an API call to my developer account rather than to the subject user. I will investigate the problem and either create a new question or update this solution with the fix I find.
Fix: Wasn't including the 'sub' qualifier to the JWT token. Fixed by modifying JWT package token generation code to include delegationEmail: user.services.google.email after scope. I don't know why he used such a long designation for the option instead of sub: as it is in the google API, but I appreciate his package nontheless.
I'm quickly becoming proficient in this, so if people have meteor-related google auth questions, let me know.
DO NOT USE SERVICE ACCOUNTS AS POSTED ABOVE!
The correct approach is to use standard web access + requesting offline access. The documentation on the api page specifically states this:
Typically, an application uses a service account when the application uses Google APIs to work with its own data rather than a user's data.
The only exception to this is when you are using google apps domain accounts and want to delegate access to your service account for the entire domain:
Authorizing a service account to access data on behalf of users in a domain is sometimes referred to as "delegating domain-wide authority"
This makes logical sense as a user must be allowed to "authorise" your application.
Back to the posters original question the flow is simple:
1) Meteor accounts google package already does most of the work for you to get tokens. You can include the scope for offline access required.
2) if you are building your own flow, you will go through the stock standard process and calls as explained on auth
This will require you to:
1) HTTP call to make the original request or you can piggyback off some of the internal meteor calls : Package.oauth.OAuth.showPopup() -- go look at the source there are more nifty functions around there.
2) Then you need to create an Iron router server side route to accept the oauth response which will contain a code parameter that you will use to exchange for tokens.
3) Next use this code to make a final call to exchange the "code" for the token + refresh_token
4) Store these where ever you want - my requirement was to store them not at the user level but multiple per user
5) Use a package like GoogleAPI this wraps up Google API calls and refreshes when required - it only works when tokens are stored in user accounts so you will need to rip it apart a bit if your tokens are stored somewhere else (like in my case)
I've some troubles whit spring-social-api for linkedin. After 60 days access token expires and I can't use it to post on timeline of access token user. There is a way to pass this mistake?
Repeating my answer from http://forum.springsource.org/showthread.php?138943-Linkedin-refresh-accesstoken :
Although I can't honestly say I've tried it with LinkedIn, you should be able to obtain the connection (via a ConnectionRepository) and call refresh() on it. In fact, this should work with any OAuth2-secured API (except for Facebook who doesn't quite play by the OAuth2 spec).
I've not tried it with LinkedIn yet, because up until recently Spring Social was working with LinkedIn via their OAuth 1.0a authentication. But it sounds like a good thing to test.
Therefore, Spring Social supports refresh of access tokens. But it is a manual effort on your part to (1) catch the ExpiredAuthorizationException, (2) use ConnectionRepository to fetch the Connection, and (3) call refresh() to update the connection.
What would be more awesome is if Spring Social were to somehow catch that ExpiredAuthorizationException for you, automatically call refresh() and then reattempt the call that triggered the exception; making it seamless for the caller. That is something I've been pondering, but there's no implementation yet.
One possible solution is an aspect. But it'd need to be configured by the developer to properly wrap whatever API binding types (LinkedInTemplate, for example) the application is using. Another option I'm thinking over is to dig down into AbstractOAuth2ApiBinding and configure the RestTemplate that it exposes to handle that exception. It's a bit iffy on whether it can work, but if it does work, then any API binding that extends AbstractOAuth2ApiBinding would automatically get refresh capability...except, again, for Facebook who doesn't play by OAuth2 rules.
Is it possible to either turn off the api explorer completely or limit the access to it?
I noticed some logs in my app that come from failed requests executed from a browser. My api is only consumed by an Android app so the only place where they can come from is the api explorer. Also the api access is limited to 1 web and 1 android client id.
Unfortunately no. The API explorer works by using the Discovery Service associated with your API, which is not actually part of your backend, so you can't specify auth or visibility for those URIs.
The list method from the Discovery service is used to generate the list on the APIs Explorer app using your app as base:
discovery.apis.list:
your-app-id.appspot.com/_ah/api/discovery/v1/apis
When someone clicks one of the APIs from the list, the full discovery document is retrieved for that apiName and apiVersion using the getRest method from the Discovery service:
discovery.apis.getRest:
your-app-id.appspot.com/_ah/api/discovery/v1/apis/{apiName}/{apiVersion}/rest
If you are looking for ways to prevent the executing of the API, check out Cloud Endpoints: Control who can execute API through API Explorer
endpoints makes auth easy and you can get the current user. You should use auth to ensure people don't mess with your private apis - otherwise people could trace what kind of post or get requests you're sending anyway - auth is always a good idea rather than trying to keep your apis secret.
If you're building a secret product and you don't want your competitor to find out, you could perhaps use some obfuscation method on the backend and on your client which makes the apis unreadable.
Also a user messing with your apis shouldn't break your database - or if it does - it should only break it for the user that was being foolish. Having logic in your client for how apis are used so that the backend doesn't break is a bad idea - the backend apis should take care of themselves and not worry about how or why they are used and who by for what purpose.