Hacker News new | ask | show | jobs
by forgetsusername 3437 days ago
>if (source.isSuspicious()

What are the attributes of the source object that we need to determine if it's suspicious? That's the hard part.

2 comments

First time user of WU sending money to Nigeria or Ghana seems pretty reasonable to me.
First time user of WU sending more than $500 to Nigeria or Ghana seems pretty reasonable to me. Let alone thousands or tens of thousands.

At the end of the day though, scammers will be scammers and victims will be victims. You can't prevent everything but you can still try harder than WU.

Really? At least you would test that the system works before handing over so much of your money to them, right?
you'd be surprised. A particularly common scam involves having the victim send 2000-2500 and the going for smaller amounts, $250-1000, on rounds 2+.

I'm not particularly well versed in romances scams but I've heard they can start in the similar range and then escalate into 5 figures quickly.

Some people have family in Nigeria or Ghana, and everyone is a first time use once. The whole point of Western Union is sending money to places that don't have a solid banking system.
Asking if they are sure it's not fraud is not necessarily preventing them from using the service. People with family in those areas will presumably know based on the wording that it doesn't apply to them. A notice that the person has to acknowledge seems a fairly innocuous defense mechanism.
Even something as simple as:

Have you physically met and personally know the person you are sending money to? Are they a family member?

If you answered no to both of these questions STOP NOW IT IS PROBABLY A SCAM. If you wish to continue please consult with the teller.

Under that criteria, 100% of eBay transactions would be scams.
If you are buying a laptop off of eBay from Nigeria and they demand that you use Western Union to pay for it there is very close to a 100% chance that you are being scammed.
I think it is valid to think in terms of expected value here, where value is negative (a loss), since there is a large volume of transactions so the law of large numbers will apply. Set a threshold of expected loss below which you will not flag transactions. Then either very large payments or relatively smaller payments to countries with a higher known percentage of fraudulent transactions will flag a transaction. You can further refine this if for example there is a relationship between observed fraud rate and transaction amount.

If a company profits off of fraudulent transactions and is aware of the fraud, and the attorney general or equivalent stops viewing this as negligence, there is an incentive for the company to encourage this fraud to increase profits.