The tone timing is in the audio sample is weird and out of spec. Minimum tone duration is 65 ms, and minimum pause duration between tones is 65ms. [1] That example has much longer tones than pauses, and there seem to be some back to back tones. The article says it's taking too long to send the data, and that's why.
If you send tones 100ms long with 100ms pauses, a conservative rate, you can get 5 digits per second.
That's about what "redial" on my phone clocks at.
If I understand what the article is saying, using DTMF tones like this does not offer anonymity for the patients on the other end of the phone call. Simply recording the DTMF tones would be enough to reveal the patient's true phone number. Perhaps there was some sort of server-side mapping scheme in use to mask the numbers and the DTMF tones were truly just for data transmission, but no such mention was made in the article.
The patient's phone number is visible to the nurse and you can see that in the app screenshot. Their goal was to mask the nurse's mobile number while having the app work offline and this solution achieved that.
In my country you can prepend '*31#' when dialling to mask your phone number.
Seems like this app could do something similar (assuming a similar dialing code is available wherever it is being used? I'd think it's a common enough feature), prepending the masking sequence to the patient's number before dialling.
DTMF was designed to interoperate with human voice and the tones were chosen on purpose to be unlikely or impossible for human voice to trigger. If there is no human voice, you don't need to use DTMF you could use any number of tones. I wonder if you could use base64 or base58 with 64 or 58 unique tones and be able to send text at a reasonable rate?
The tone timing is in the audio sample is weird and out of spec. Minimum tone duration is 65 ms, and minimum pause duration between tones is 65ms. [1] That example has much longer tones than pauses, and there seem to be some back to back tones. The article says it's taking too long to send the data, and that's why.
If you send tones 100ms long with 100ms pauses, a conservative rate, you can get 5 digits per second. That's about what "redial" on my phone clocks at.
[1] https://www.etsi.org/deliver/etsi_es/201200_201299/20123502/...
If I understand what the article is saying, using DTMF tones like this does not offer anonymity for the patients on the other end of the phone call. Simply recording the DTMF tones would be enough to reveal the patient's true phone number. Perhaps there was some sort of server-side mapping scheme in use to mask the numbers and the DTMF tones were truly just for data transmission, but no such mention was made in the article.
The patient's phone number is visible to the nurse and you can see that in the app screenshot. Their goal was to mask the nurse's mobile number while having the app work offline and this solution achieved that.
In my country you can prepend '*31#' when dialling to mask your phone number.
Seems like this app could do something similar (assuming a similar dialing code is available wherever it is being used? I'd think it's a common enough feature), prepending the masking sequence to the patient's number before dialling.
Fax is still acceptable in parts of healthcare for a reason — privacy under low-tech constraints is an actual requirement.
Dial-up was slow, but at least the internet still felt human.
Fiber gave us speed, not soul.
Sometimes I miss yelling “Corp Por” into the TUBE — back when the screen wasn’t a window, but a passage.
DTMF was designed to interoperate with human voice and the tones were chosen on purpose to be unlikely or impossible for human voice to trigger. If there is no human voice, you don't need to use DTMF you could use any number of tones. I wonder if you could use base64 or base58 with 64 or 58 unique tones and be able to send text at a reasonable rate?
They could have used modem standards. Bell 103 standard is 300 bit/s with frequency shift keying (FSK).