Abating Telephone Nuisance Calls
Due to the proliferation of telephone technology, at present there is no reliable way of caller id This has allowed rampant telemarketing, both legitimate and fraudulent. I believe that it would be possible to create the equivalent of an telephone captcha on a device that could be wired into the end user's phone system at telco’s access point, and be programmed through wifi. Such a device would have a simple web interface over wifi, and be configured by the owner. This could be done either as a box at the subcriber, or as a telco service at hte exchange.
I get frequent calls, usually from the U.S. with a recorded message, “Stop what you are doing. How would you like to make ten thousand dollars a month…”; “This is a time limited offer, there is no problem with your credit card”; Or sometimes a non-recorded message. Or even an aluminum siding salesman.
I have complained to the CRTC. They have told me to tell my service provider. I have complained to Telus. They say nothing can be done. I have looked at the problem, and it is not trivial. There are companies working on hammering out a set of standards. This will likely not happen quickly. Twenty minutes after writing this paragraph, I received a report from Bell Canada et al due to my complaints. Summary: It’s too soon, too hard, too expensive to even comment on it at this point. This tells me that there is likely at minimum a 5-10 year market for this device.
I envision a moderately smart box that screens your calls for you. It basically is about as complicated as a hybrid answering machine and dialup modem.
I’m not the first with this idea: Failed project of smaller scope on kickstarter: (Links not allowed here.)
details an Arduino interface that can make/intepret telephone DTMF tones. It also brings up interface requirements of the POTS, and possible legal issues.
I have in mind a box, probably based on Raspberry Pi or Arduino.
Daughter cards:
Operation Cycle -- Incoming call.
Operation cycle -- outgoing call.
This system has very flexible configuration: I have gotten a few spam calls that ID’d as local exchange. Returning the call got either a ‘not me’ or not in service. If this becomes common, then the user can require a passcode for all incoming calls.
More about Challenges.
Challenges can be of various forms:
In the first three cases the device can pick a question randomly from a list; users are encouraged to dictate their own questions as it challenges voice recognition software. The process can be made harder for bots by ‘augmenting’ the recording with some amount of static, hisses, children playing, dog barking, etc. (People are much better than machines at dealing with signal noise)
Configuration:
This is a first sketch of the configuration. In general one of the problems of the Internet of Things, is that most devices are neither secure nor readily upgradeable. For this reason, I would propose that configuration be done on the manufacturers web site, then the local browser be used to program the Call Screener. This requires that the user register his device, and provide a way to be contacted. This may also be done through a smartphone app.
On start, the configuration app fetches the model/serial number of the device. This allows multiple versions to be maintained. The current firmware version and current config is also fetched.
Options:
Area code Whitelist. I find that most problem calls originate in the U.S. So at this point I would allow all the area codes in western Canada.
Area code Blacklist. For users who have more eclectic regular callers, they may want to work by disallowing codes instead. Note that modified versions of this would be required in areas where country codes are in common use. It may be desirable to have multiple lists.
Phone number blacklist. Based on caller id. This can be spoofed. This configuration page could also be used to set policy on “Private Caller” “Unknown” or cases where the number or state only is present with no further information. On on those ones where Vcode is show as caller ID.
Phone number whitelist. These are numbers that go through automatically, without audio captcha.
Both white lists and black lists are not pure black and white. You may say tnat blacklisted area codes/numbers have more limited hours/harder challenges.
Hours. One or more lists of times. Times can be in terms of weekly schedule, holidays etc.
Captcha questions. Recordings of questions, and their answers.
The remaining configuration is an action plan:
And one of these:
Thus you can still have a private life on a business phone.
Sample Action Tree
Introduction. “We are call screening. If you have a code, enter it at any time. If you are a business customer, this months’s code is below the logo on our website, or you can use the one in our latest flyer.
Examples:
We have a business that normally closes at 8 p.m, but we don’t want cold calls after 5, but are willing to have current customers call to arrange a pickup.
So part of our email signature is this month’s current customer code. When that code is punched in, and it’s not yet after 8, we get the business phone ring pattern. If after 8, goes direct to voice mail. If it’s a friend’s code, it comes through it gets passed through with a family ring. If no code is punched after a suitable period, the device starts an action tree.
The configuration is complex. It must have an easy way to both set it up and test it. It should make a good diagram of it’s configuration.
One of the ways to configure it may be to posit scenarios:
Markets:
There are at least three different markets for this sort of device.
The threat of marketing this as a consumer item may be enough to convince telcos to add it at the exchange level.
Prototyping
Version 1 to determine feasibility and interface could likely be done on an older PC with serial out, and a modem. Version 2 on a cheap single board computer. Production version would be a single board.
Power. While phones continue to work when power goes out, Telcos will not be amused by even small parasitic (to their view) loads on the system. That said, the power demand for something like this should be very small.