3 ms·

Can anyone explain why this is hard, technically?

by arohner 14y ago

anonymous1019 14y ago
Sure. If you plan on implementing this as some sort of end-user device that would be hooked up to a phone handset or a software "app" you would install on a smart phone, then all you've got to work with is caller ID. Caller ID can be blocked by the caller (e.g., by dialing *67 first) and spoofed, including the purported outgoing number. In fact, VoIP systems like Skype have made spoofing caller ID and now even ANI, a toll network analog of caller ID, trivial. So even if you keep some sort of constantly updated database of numbers used by robocallers, you are still relying on the robocallers 1) not blocking outgoing caller ID and 2) not spoofing the numbers of legitimate users resulting in them getting blacklisted.
- hollerith 14y ago
  >VoIP systems like Skype have made spoofing caller ID and now even ANI . . . trivial. Is there any way for hardware connected to an ordinary phone line to distinguish between an incoming call from a VoIP system versus an incoming call from an ordinary phone line?
andreasvc 14y ago
For the same reason spam is a hard problem. You'd like to maintain an open system, but unfortunately there are cheap ways to exploit that.
- cesium62 14y ago
  Not completely true. Spam is relatively easy to block because you have a complete copy of the communication to work with and not just the envelope. In the spam world, honeypots are set out. If an email is sent to a honeypot (a fake address), then it is, with high probability, spam. If the email is sent to multiple honeypots, you can be sure it is spam. If a variety of humans mark the email as spam, you can be reasonably sure it is spam. In principal, you can now compare some signature of an email with known spam messages. These techniques can be used with phone calls, but there is less data to work with. The approach that applies a turing test to each incoming phone call will filter out robocalls nicely. And it increases the costs of humans calling enough to filter out many of them. For persistent humans, I like the approach of having the sender send a token payment to the receiver which the receiver can easily refund.

Can anyone explain why this is hard, technically?

by arohner 14y ago

Can anyone explain why this is hard, technically?

Sure. If you plan on implementing this as some sort of end-user device that would be hooked up to a phone handset or a software "app" you would install on a smart phone, then all you've got to work with is caller ID. Caller ID can be blocked by the caller (e.g., by dialing *67 first) and spoofed, including the purported outgoing number. In fact, VoIP systems like Skype have made spoofing caller ID and now even ANI, a toll network analog of caller ID, trivial. So even if you keep some sort of constantly updated database of numbers used by robocallers, you are still relying on the robocallers 1) not blocking outgoing caller ID and 2) not spoofing the numbers of legitimate users resulting in them getting blacklisted.

hollerith 14y ago
>VoIP systems like Skype have made spoofing caller ID and now even ANI . . . trivial. Is there any way for hardware connected to an ordinary phone line to distinguish between an incoming call from a VoIP system versus an incoming call from an ordinary phone line?

andreasvc 14y ago

For the same reason spam is a hard problem. You'd like to maintain an open system, but unfortunately there are cheap ways to exploit that.

cesium62 14y ago
Not completely true. Spam is relatively easy to block because you have a complete copy of the communication to work with and not just the envelope. In the spam world, honeypots are set out. If an email is sent to a honeypot (a fake address), then it is, with high probability, spam. If the email is sent to multiple honeypots, you can be sure it is spam. If a variety of humans mark the email as spam, you can be reasonably sure it is spam. In principal, you can now compare some signature of an email with known spam messages. These techniques can be used with phone calls, but there is less data to work with. The approach that applies a turing test to each incoming phone call will filter out robocalls nicely. And it increases the costs of humans calling enough to filter out many of them. For persistent humans, I like the approach of having the sender send a token payment to the receiver which the receiver can easily refund.