Came from recaptcha v2 background with my own bot detection logic. I want to try out the new bot detection score feature of v3.
I want to protect my search page from being abused by bots trying to query the DB every 0.5 second.
Search result page accepts url query parameter to allow users to book mark the page. For example mydomain.com/findallcities?country=xxx&province=yyy&populationmin=5000
Based on my understanding v3 flow works like this
Client calls recaptcha JS function and get a token
Client append the token some where in the form so that server can validate the token later
Form submit and server receives the token
If server gets no token then reject. otherwise validate the token to get user score and act accordingly
So the question comes. How can I tell if its a bot that is hitting mydomain.com/findallcities?country=xxx&province=yyy&populationmin=5000 with a Get request or if its a legitimate user clicking on a bookmark? There is no token to verify and even if there is its probably an expired (used) one from before.
Related
This is my current understanding of recaptcha (using v2 invisible)
We load the api.js script onto our site
We add data attributes to the button
User clicks button
A listener somewhere in the api.js script fires because it's listening for an event on a tag with those data attributes
This is where it gets fuzzy and I start guessing:
api.js gathers browsing info from the user's cookies and information about how they interacted with the site. Based on this it determines how likely you are are to be a bot and if you are below a certain threshold it gives you a test. Whether you pass the test then gets further factored into your score and all of that gets encoded into a token, which we receive in our callback that we specified on the button's data attribute.
We pass this token to the back end with the rest of our form
From the backend, we make an API request to Google to convert the token into usable information about whether the user passed or failed.
At this point I get confused about why this wasn't just what the api.js script returned in the first place. Does this step only exist in order to give Recaptcha information to further improve it? I just don't understand why this step is here, unless I'm misunderstanding what is going on earlier in the process. Am I getting these we steps wrong? Thanks.
The whole point for captchas is that your server (instead of client in the browser) can verify that the (HTTP) request it received was generated from a real person's actions, when interacting with your application.
This is why your client sends a recaptcha token to your server and your backend consults with the captcha provider about this token and receives trusted information about the original client. In this scenario, your server does not trust the client, so it receives only a token from it. Then it communicates with the trusted captcha provider server-to-server and validates that the token it received from the client is valid and the user behind it is legitimate.
If your client sent the original response from the captcha provider to your backend server, there would be no way for your server to know whether this was a legitimate response from the captcha provider, or a fake one from the client.
I am using Laravel as my backend together with Sanctum which generates personal access token for mobile users. For my mobile application I am using flutter.
To authenticate users they login with their username/password and get a personal access token in return. This works but requires a user to login every time they open the application again so I did what most tutorials suggest which is saving the token on the mobile device using shared preferences/secure storage.
Now comes the question how do you invalidate a user when you remove their token from the backend? On initial login it appears everything is still fine because like in most tutorial I check for the existence of a token. After that whenever I want to make a request which uses the token I obviously run into problems because it not longer exists on the backend.
Most tutorials/guide suggest saving the token and using that a reference to see if the user is logged in or not but this seems flawed because it gives the false impression you actually have a valid token.
My guess is this can be solved by always performing a heartbeat/ping action to check if the current token is valid and if not send them to the login screen instead of simply checking for the existence of the token.
Thoughts on this?
I can suggest a hack or trick here in every launch of the app you can send a request to an API to check if the user's token is valid or not and if it is valid then you can continue the app otherwise force the user to login and generate new token this way your app will be secure via server / API.
For this, you can store the user's secret token in the database and check it via HTTP API call and send a response from the API accordingly and check the response in app and do the next operation according to the response you get.
I don't know if this is a great way of doing this job but it is a kind of hack/trick to achieve what is needed.
Thanks
I want to ask you how does sites like Facebook etc. handle user authentication on the page refresh? I wonder why I don't see any loader when I refresh the page and I'm on the protected route (e.g. my profile settings).
When I create an application I use following authentication flow:
Load the application
Show authentication process loader to the user.
Get user session token from the cookie.
Send authentication request with the token to the backend.
Validate the token on the backend.
Send a response to the client with the status code 200 if token is valid or status code 401 if token is invalid.
Receive a response on the client side and hide the loader.
Show protected route to user if he is authenticated or redirect to the login page when he is not authenticated.
So with such a flow initial loader is a necessity, how facebook is able to omit that?
and welcome to Stack Overflow.
I'm not sure your question is valid - pretty much every page on Facebook requires you to be authenticated (not just the profile page), and authenticated access is a common feature of pretty much every major website; they rarely show loading bars.
We don't know exactly how Facebook is built, but it's a fair guess they use React; you can read up on how to manage authentication on React here.
Your flow is predicated on the assumption that there is an "application" to load, and that the authentication request will take substantial amounts of time. If your token validation is fast, this assumption is probably incorrect. You do see "loading" animations on React apps - if you scroll down your timeline on FB really fast, you see the circle; this is probably partly network, partly retrieving the posts.
I figured it out. It's possible because authentication can be done on the server side using cookies. They validate user token when user enter the page and after this they render desired version of page (different for authenticated and different for not authenticated user).
My project has the requirement to access the yammer data using the given REST API using server side script(mainly PHP) and not involve a client side login using yammer's OAuth dialog.
I have gone through this document:
https://developer.yammer.com/docs/oauth-2
but this says, we requires user interaction.
What I wanted was can I generate a client_id and client_Secret to further generate access token to make API call out, but in all these processes I only use the authenticated users username and password in my server-side script.
Can anyone suggest a solution or is a client-side interaction required by design?
Thanks in advance!!
You have to have a user authorize the application at least once. This is just the nature of the OAuth implementation and you can't work around it. Having users go through the OAuth flow is considered a best practice.
If you have an OAuth token for a verified admin of Yammer, you can use impersonation to get tokens for end users without them interacting with the OAuth flow.
The below from Microsoft blogs might help you & added source at the end of answer.
Obtain a Verified Admin token for your application in one of the following 2 ways
a. Create the app with a Verified Admin account and then in the app’s Basic Info page, click “Generate a developer token for this application.” Note that you’ll need to use this app’s info in the JS SDK and any subsequent calls.
b. Use the process outlined at https://developer.yammer.com/docs/test-token with a Verified Admin account to get an OAuth token for that VA account. Note that you must use the app info used to generate this token in all future steps.
Obtain the current user’s email address in the server-side script.
Using the VA token obtained in step 1 to authenticate, pass the user’s email address to our Get User by Email Address endpoint documented at https://developer.yammer.com/docs/usersby_emailjsonemailuserdomaincom, and then process the response
a. If the call to the API endpoint returns a 200 OK response, first check the “state” field to make sure the user is “active” and if so, store the “id” field that’s returned and go to step 4
b. If the call returns a 404 or a state other than “active,” direct the user to finish creating and activating their account however you like.
Once you have the user’s ID, you can pass it to our Impersonation endpoint to obtain a pre-authorized OAuth token for that user. This endpoint is documented at https://developer.yammer.com/docs/impersonation and must use the VA token obtained in step 1 to authorize the call, and the consumer_key of your JS SDK app.
You now have an OAuth token for the current user. When generating the code being passed to the browser, have the client side JS SDK code first call yam.platform.getLoginStatus and if there’s no active session and you have a token from step 4, pass that token to yam.platform.setAuthToken($tokenFromStep4, optional_callback_function_if_desired(response)). If you don’t have a valid token, direct the user to finish setting up their Yammer account.
Continue making JS SDK calls as you normally would, without needing the user to authenticate.
Source: https://blogs.technet.microsoft.com/askyammer/2016/11/04/preauthorizing-the-yammer-js-sdk/
App was developed in Web API. We are using AntiForgery Token validation for all the POST calls in xyz.com. Registered xyz.com users get email alerts for the contents they are signed up for. Users click item from email to view the content details. On click of item first click event is being saved in SilverPop and silverpop redirects to actual content in xyz.com. Details page is being loaded with out any issues after authenticating the user based on few query string parameters. Issue is when user make a post call(to saving for later, book mark etc) App is throwing Antiforgery token mismatch error. Sorry for bad English and long question. Strange part of this issue is We are unable to replicate this error and its not occurring consistently. We have two server and 1 load balance hosted in AWS not using sticky session enabled. Using Forms cookie authentication. finally, this issue kept occurring though we run with 1 server. Thanks for reading and appreciate if any one could help.
*User must be Authenticated to validate CSRF token ? This xyz.com makes few api get and post calls.