Ask HN: Service for detecting spam signups?

3 points by TimJRobinson ↗ HN
At our startup we've recently started seeing a surge in spam / fraudulent users signing up to abuse the service. We have a captcha on our registration form but it isn't enough.

I was wondering if there's a service out there where you can submit someones username + some other details such as name, email etc and it would give a rating on how likely that signup is to be a spammer or a legitimate user.

Does something like this exist?

It's fairly easy as a human to look at a new users details and determine whether they are likely to be a spammer or not (random character username / email / names or a bunch of accounts signed up in quick succession with similar details) but we don't have the resources to do this and it would be great to have an automated way to do this.

8 comments

[ 4.7 ms ] story [ 32.1 ms ] thread
One idea might be to use an API like Ark (https://ark.com/) or Clearbit (https://clearbit.com/) to see if the email has any social accounts associated with it, and if it doesn't then label it as a high likelihood of a fake signup.
Lots of our users use their business address (B2B service) and that wouldn't work.

On the one hand I wish a service could filter mailinator, 10-minute-mail, trash-mail addresses (each have hundreds of domain aliases), on the other hand I'm using anonymous email myself sometimes so I'm glad it doesn't exist.

Off the top of my head I remember http://www.stopforumspam.com/ exists - which is something I use n my on blogspam-detector http://blogspam.net/

But I suspect that you'd be better placed to look at the patterns yourself, if you were to setup some simple node.js/dancer/service to test an IP/email/name combination and return a "yes/no" result you could then experiment with different metrics.

One thing I do on my sites is block IPs. So if bogus-user-bob signs up from 1.2.3.4 then no other user will be permitted to sign up from that IP. I realize that proxies will allow this to blacklist too many remote users, but in practice I haven't had complaints.

https://akismet.com/, known for identifying blog on Wordpress blogs, has an API.
Our API is entirely focused on identifying comment spam; we don't have anything that would be useful for identifying spam signups.
I'm aware of a load of hosting companies using the MaxMind service, you basically push them a load of information for a transaction and they reply with a risk rating.

I've use it in the past a few times and it was fairly accurate at capturing the fraudulent traffic.

I have used binary classifier to detect Ham and Spam emails i am pretty sure we can apply some sort of classifier here as well i am not sure about the exact parameters but if we have enough sample data we can surely write a classifier for this problem as well. If you have sample data for spam signups and regular signups tweet me at @alihaider907 can guide you in this regard