Comment by js2 | Hacker News Reader

Comment by js2 | Hacker News Reader

I've been getting this call[1] literally every month for about a year now. It goes straight to voicemail because their dialer can't get past my voice captcha. I tried blocking the caller id but they keep using a different last four numbers. I've now configured the call flow so that this entire exchange now has to pass the captcha to even get to voice mail.

What I'm saying is, this proposal can't come soon enough.

1. Transcription, for your amusement: Hi this is Carolyn calling from reliable resource communications. Reliable resource communications is a telecommunications service used by other companies to notify consumers on their behalf. I have made numerous attempts to reach you regarding an entry form that was filled out in your name within the last 12 to 18 months to receive a new car. Now this will be my final attempt to notify you that your name was pulled and you are going to receive one of our top three major prizes. It will be in your best interest to give me a call back at soon as possible. My number is 984-292-1515 at extension 3:21. We are not a telemarketing agency nor timeshare and this is not a cold call please do not ignore this message. I'm very aware of the do not call list but wouldn't be calling unless someone actually answered. This is a time sensitive matter and I do look forward to hearing from you. Once again congratulations my name is Carolyn.

> this will be my final attempt to notify you

It's always the final attempt/notice. Every time. I've received dozens of "final notices" that my 22 year old car's factory warranty (10 yrs, 100k miles) is "about" to expire.

Get this one all the time. I did have a car warranty that expired (a while ago), so I always wondered if its just a stock message, or if someone actually sold the data of my car warranty.

In a similar vein: warnings for your 'last chance' to apply for house insurance. As someone who likes to keep physical mail to a minimum, all of the above are rather annoying.

I've been getting a bunch of sketchy post cards about extended warranties using wrong language about my warranty just expiring on a car I purchased used (9 years old, 50k miles, 3rd owner according to CarFax). I've been filing postal inspector requests on their bulk mail code every time, but no response so far. That particular sorta scam is not limited to phone calls.

Oh, you are handling those calls all wrong. On the odd times I do actually pick up I choose from a handful of possible replies:

1. I am in prison for vehicular manslaughter and how did you get my burner cell phone number? 2. I was in a horrible accident in that car and my life is ruined, thanks for bringing it up. 3. The more general "I only have 6 months to live" I use as a catch-all for ending any of these calls: sure I am lying, are you bold enough to call me on it?

#3 makes me think I should talk to my wife about being ready for me to hand off the phone in that situation.

This call is about 3 out of 4 of all phone calls I get.

Years ago the majority of my emails and phone calls became spam. Now it's the great majority.

The majority of people who knock on my door aren't yet religious proselytizers, but it's getting close.

What's this voice captcha you use?

Anveo Call Flow. Numbers not on a whitelist get a recording "press 1 to be connected" which repeats three times a couple seconds apart. If they don't press 1 they get sent to voice mail. I also have a blacklist that gets a captcha which says "press 1 to leave a voice mail" and hangs up if they don't.

On the one hand, that's so smart I can't believe it's not widely used by everyone.

On the other hand, if it was that common, there would be a market for systems that defeat it for the spammers.

If bypasses were common (mostly just sending "1"), then a configurable number you could cycle would work fine. Everyone would have a different number and you only get 3 tries.

Anything more complicated like actually following the voice prompt isn't going to happen soon.

voice recognition is already to the point where it can recognize numbers. That we are limited to one of 12 numbers (including * and #) before it gets too hard for humans limits the problem scope to the point where it is a relatively easy problem.

Of course this assumes enough people follow the pattern of press some number to continue. Every time the pattern changes the cost of understanding it and fitting it into their system goes up. In the end though it isn't a scheme we can win: AI can successfully follow more complex processes than humans can in this limited space - including random failure to follow the instructed process if it is too complex.

It can be done by asking some obvious question for humans:

"This is a anti-robocall system, please answer the question to complete the call. A mother cat and her three kittens are sleeping in a box. Dial the number of animals in that box."

The number of kittens change randomly and maybe the system can put some pillows, blankets and other non-animals in the box too, making automated processing harder. AI will get there, sure, but this will make things expensive for the attackers.

Your example is a trick question that will catch some real humans.

Some might consider that a feature.

I have run asterisk for 15 years now, everyone in the house has a 1 digit extension number (that way we can have private voicemail) I haven't had a robocall in years

I've always thought that binding robocallers in a contract that they must accept to speak or leave a message would be fun. Notify the caller that you charge a fee that's your state's maximum small claims award and add that the fee to be waived at your discretion for legit business. The big problem is spoofing and overseas calls, but I'd bet you could nab a few cases.

I'm thinking something like, "All unsolicited calls concerning the sale, lease, rental, or exchange of goods and/or services, regardless of past contact between calling parties or their third party affiliates, will incur a $5000/hr convenience fee, minimum billing for 1 hour, to be paid to _your name_ at POBox ... in full within 90 calendar days of the current date. This fee may be waived at _your name_'s discretion. Press 1 to accept or hang up to decline."

Bonus points if you fit the fee discussion in before the call center switchboard hand-off to a live operator.

I actually got so fed up with robocallers, and really people just calling me in general that I looked into setting up a 900 number. If people really want to intrude and take up my time, they can compensate me for it.

Unfortunately, 900 numbers apparently went away in the 2000s. Such a shame, it really would have solved the problem quite efficiently.

Wasn't there some kind of scam people used to pull where they'd add a prefix to a cell phone's outgoing calls, such that it bounced through a paid service relay? If I recall, someone would nab your phone, add the dialing prefix to the settings, and then return the phone to you. From that point on, all outgoing calls would actually be calling a 900 number, which then just called the outgoing number. The phone's owner would then get caught with a massive bill from the 900 number. It was an updated version of one of those original 'boxes', maybe a modified brown box? I forget all the phreaking terminology now, but I wonder if there's another olde timey phone phreaker trick that could be exploited.

For the captcha, do you have any proposal for what to do when the caller is legitimate but doesn't speak English?

If someone is legitimately calling me, they need to be able to speak English, or I need to take a few courses in a foreign language first

I'd assume he's not expecting foreign callers

If parent only speaks English, I don't see the problem.

As a proposal for wider implementation, you could just select the languages you speak to be valid languages for the captcha. Seems like a feature rather than a problem.

Are there any other, maybe more consumer-friendly, anti-robocall tools?

Wasn't her name Carolyn?

Reply on news.ycombinator.com