Idea: Human screenning
Paul Bohme
paul at bohme.org
Tue Jan 16 12:19:10 CET 2007
Ole Tange wrote:
> On 1/15/07, Paul Bohme <openmoko at bohme.org> wrote:
>> Dave Crossland wrote:
>> > On 15/01/07, Gervais Mulongoy <gervais.mulongoy at gmail.com> wrote:
>> >> Heh, until the phone spam operators start using basic voice
>> >> recognition and
>> >> to defeat the simple riddle :p
>> >
>> > Spammers don't do email address de-obfuscation because it takes too
>> > much processing time; I can't see them doing this in practice :-)
>>
>> Unfortunately, with zombie bot-nets being what they are, spammers have
>> essentially infinite CPU power at their disposal. Whether the
>> investment to get through a single obstinate node is worthwhile, look
>> for any mechanism short of a pure white-list to be eventually overrun.
>
> Could you elaborate on that?
>
> To me it seems that if you can solve riddles then you are fairly close
> to making a computer that passes the Turing test. Especially if the
> riddles are not of the same form: "What is ten plus hundred?" "Enter
> the last five digits of my phone number" "Enter yesterday's date"
>
> First of all you need to solve speech to text. Then you need to parse
> and understand the sentence. Both problems have proven to be really
> hard problems. Throwing a botnet after this is not going to solve it
> if the algorithm is simply not there.
Thus the 'Whether the investment to get through a single obstinate node'
comment above - acknowledgment that cases like this probably won't
matter enough to warrant the effort on the part of the unwanted caller.
As a more general note, many suggestions have been put forth in the
email realm to incur some CPU cost for sending an email - these do fall
readily enough given enough CPU power.
If someone were to arrange a central call-filtering mechanism to kill
off these kinds of calls, count on someone else circumventing it.
Discussions about algorithms and problems are all well and good, but
this is money.. ;-)
-P
More information about the community
mailing list