Live data from Hacker News

Tell HN: Automatic fraud detection is making my life hell

news.ycombinator.com

281–290 of 406 posts

Re: Tell HN: Automatic fraud detection is making my life hell

#281
post #137

Earlier quoted context omitted.

The problem with many of these examples is that 99% of the time, it is a sign of fraud, and 1% of the time it’s a false positive. > If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. I’ve been using this heuristic (along with VPN and IP geo lookup) when screening job candidates after a massive influx of developers outside the US applying for US-only remote roles. I discov…

My primary phone number is a Google Voice number, and I am entirely legitimate. Just curious, how often are other people filtering with this kind of criteria? I have been considering migrating away from GV for unrelated reasons, but if that sort of thing automatically makes me less attractive when looking for gigs then I'd like to prioritize actually doing that.

I primarily use Google Voice, especially since I’ve been out of the USA for a while. I haven’t had any major problems, but I have noticed things getting a little more difficult. I was recently denied a savings account at a bank with whom I already have a credit card because they couldn’t text my Google Voice number. They offered to try calling me at a different phone number that’s known to be associated with me, I couldn’t just give them a different number, but I hadn’t uses their proposed number in nearly 20 years.

Re: Tell HN: Automatic fraud detection is making my life hell

#282
post #261

Earlier quoted context omitted.

> The problem with many of these examples is that 99% of the time, it is a sign a fraud, and 1% of the time it’s a false positive. That’s the key. It may very well be the wrong business decision to care about this 1%.

Idk, if your aim is to find the "best talent", then what's the chance that they stumble along and you treat them like shit? That's what's going to happen when you say "out of this other group, this 99% of people who I didn't want anyway, many of the Google Voice people were fraudsters". Same thing for asking people to reverse a linked list on a whiteboard, or getting them to re-do their résumé, but in your HTML form…

With most interview processes, your aim is to have a high degree of certainty that you will find someone in the top 1% or so of people, not to find the absolute best person. Given that, arbitrary filters that save your time are very much worth it.

Re: Tell HN: Automatic fraud detection is making my life hell

#283
post #254

Earlier quoted context omitted.

I'm confused about what such applicants hope to gain by this. What's their business model? They're going to have to give you a SS# or EIN (if they're a self-employed consultant) before you can send them a paycheck, right? And the Social Security Administration has a website where you can verify SS numbers. So what's the play here?

https://leaddev.com/team/beware-fake-job-candidates

This is a good description of the problem - kinda from both sides. Thank you! But based on this, wouldn't the right defense be to insist for some camera feed?

1) Most phones have a camera now - besides the plausible excuse of the candidate's computer not having a working one. 2) When no camera is working, it usually shouldn't be a big problem to postpose the interview for a few days. The time for a camera solution to be procured by the candidate. 3) For an extra test of the candidate being able to engineer their way around a broken or missing camera in less than a week.

Re: Tell HN: Automatic fraud detection is making my life hell

#284
post #181

Earlier quoted context omitted.

So you wont hire anyone who uses google fi? That's pretty lousy.

Google fi phone numbers are not VOIP they operate as a MVNO. Google voice numbers are VOIP.

It's been a while since I worked with this, but when Google Fi started, their numbers would often show up as Google Voice in carrier lookups, sometimes only for the first few days of service though.

Re: Tell HN: Automatic fraud detection is making my life hell

#285
post #137

Wrong assumptions programmers make about fraud prevention: -- A mobile phone number uniquely identifies a single person. -- Every person has a mobile phone number and they only have one mobile phone number. -- If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. -- Every person always has their mobile phone handy and it is always able to receive calls and SMS messages under…

The problem with many of these examples is that 99% of the time, it is a sign of fraud, and 1% of the time it’s a false positive. > If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. I’ve been using this heuristic (along with VPN and IP geo lookup) when screening job candidates after a massive influx of developers outside the US applying for US-only remote roles. I discov…

- Different signals mean different things in different contexts

VOIP phone numbers can indicate an increase in the chance of fraud in some contexts, and be almost meaningless in others

Re: Tell HN: Automatic fraud detection is making my life hell

#286
post #200
post #137

Earlier quoted context omitted.

The problem with many of these examples is that 99% of the time, it is a sign of fraud, and 1% of the time it’s a false positive. > If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. I’ve been using this heuristic (along with VPN and IP geo lookup) when screening job candidates after a massive influx of developers outside the US applying for US-only remote roles. I discov…

You could use identity verification: https://plaid.com/products/identity-verification/ https://stripe.com/identity https://www.id.me/

Or you know, your brain

And don't accept candidates without video chatting first, or who behave weirdly in interviews

Re: Tell HN: Automatic fraud detection is making my life hell

#287
post #148
post #141

Earlier quoted context omitted.

All these can be the wrong assumptions for sure. Working in a space with some fraud though I can tell you the majority of users verify fine and there ate only the few percentage that don’t for any number of reasons some what you have given above. For smaller vendors though things to consider are - a charge back can come up to 6 months later. A loss of that is not only a loss of funds but a charge back fee - too many…

The problem is that it's not acceptable for online providers to converge on making life miserable for "a few percent" of people. This is a market failure: to save the cost of a few bucks, huge costs are imposed on these individuals. The answer is to have some mechanism whereby people who run into these issues can pay ( once ) the small cost of being validated in an alternative way (like, actually talking to a human a…

>The answer is to have some mechanism whereby people who run into these issues can pay (once) the small cost of being validated in an alternative way (like, actually talking to a human and explaining what's going on, which is how these issues got solved in meatspace originally)

The answer is to have electronic money accounts and transfer services be operated by the government, and to make it the government's problem to go after criminals rather than have the businesses left holding the bag.

That way the business is not incentivized to discriminate, as long as they get the money through the government money transfer service, they are guaranteed it as if they received cash.

Corollary is that you also need a law that guarantees the right for everyone to have an electronic money account that can send and receive money and that no government can take that ability away from you at any point in time.

And the government has to operate an identity verification API. And again, the onus is on the government to go after criminals committing fraud.

Once you put the onus of fraud or damages on a business, then every business will obviously start discriminating to minimize those costs.

Re: Tell HN: Automatic fraud detection is making my life hell

#288
post #137

Wrong assumptions programmers make about fraud prevention: -- A mobile phone number uniquely identifies a single person. -- Every person has a mobile phone number and they only have one mobile phone number. -- If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. -- Every person always has their mobile phone handy and it is always able to receive calls and SMS messages under…

The problem with many of these examples is that 99% of the time, it is a sign of fraud, and 1% of the time it’s a false positive. > If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. I’ve been using this heuristic (along with VPN and IP geo lookup) when screening job candidates after a massive influx of developers outside the US applying for US-only remote roles. I discov…

Seems to me you are filtering against agile, tech-aware candidates - prefering less agile, less tech-aware candidates. These are your false positives. A time saver for sure, but perhaps not optimal for filling the positions.

As opposed to, for example, demanding a video feed (and giving time for an alleged engineer to engineer their way past an alleged broken camera.)

It used to be that demanding a photo or video was frowned upon as being a bit too easy to use to filter on race or gender. But I guess not anymore.

Re: Tell HN: Automatic fraud detection is making my life hell

#289
post #141

Earlier quoted context omitted.

All these can be the wrong assumptions for sure. Working in a space with some fraud though I can tell you the majority of users verify fine and there ate only the few percentage that don’t for any number of reasons some what you have given above. For smaller vendors though things to consider are - a charge back can come up to 6 months later. A loss of that is not only a loss of funds but a charge back fee - too many…

> only the few percentage that don’t Few percent of transactions for a payment processor means a few billion transactions. Visa on its own processes ~200B transactions a year. That's not a great threshold.

It's a trade-off though, right? Presumably it's better for the company to lose out on a small percentage due to false positives than to allow them through and subsequently incur a larger percentage of false negatives. They can aim for obtaining those transactions without incurring extra fraud but it has to be worth the cost and risk of the changes to the system.

Re: Tell HN: Automatic fraud detection is making my life hell

#290
post #141

Wrong assumptions programmers make about fraud prevention: -- A mobile phone number uniquely identifies a single person. -- Every person has a mobile phone number and they only have one mobile phone number. -- If a person's mobile phone number is associated with VoIP or Google Voice, that indicates fraud. -- Every person always has their mobile phone handy and it is always able to receive calls and SMS messages under…

All these can be the wrong assumptions for sure. Working in a space with some fraud though I can tell you the majority of users verify fine and there ate only the few percentage that don’t for any number of reasons some what you have given above. For smaller vendors though things to consider are - a charge back can come up to 6 months later. A loss of that is not only a loss of funds but a charge back fee - too many…

Nobody contests that it's convenient. The question is whether it's technically correct. And the point is that it's not. Just because it's easy doesn't mean you found a solution.
Post reply on HN