Pindrop Security offers Phoneprinting a voice recognition and data identification application. The site explains how it works to identify fraud callers. Callers that may use different phone numbers but in 30 seconds of phone talk it can match the voice to a data base linking Customer voice prints to various data elements associated with that customer.
Pindrop also has an Intelligence Network Network metrics are cited but none that specify how many fraud associated voice prints they have in the data base. Prindrop's data base is huge!
Who else has a huge voice print data base?
Verint
Some (that mysterious somebody) speculate the NSA records voice prints. Maybe it is just the voice prints of the known bad guys? Voice prints might be considered meta data related to a phone conversation? If all the spoken words are reduced to some algorithm that is no longer natural language speech and that unique identifying result is associated with a telephone number, date, time, etc. Then it is a data element that can be compared to subsequent voice print analysis to identify matches based of the voice print algorithm match independent of the telephone number. Certainly a useful intelligence tool. Might the voice print analysis algorithm depend on an actual test comparison of speech to data base speech or is real time speech reduced to a non speech algorithm then compared to the non speech algorithm data base?
Agnitio appears to be a major provider of voice print solutions.
https://www.rt.com/news/195552-voiceprint-harvesting-millions-privacy/
http://nworeport.me/2014/10/13/millions-of-voiceprints-quietly-being-harvested/
This could be another opportunity for business to monetize their data base by selling the voice print data? Perhaps it could be a business model to somehow harvest voice prints? Everything said after "hey Siri" goes to Apple. Where to then?
No comments:
Post a Comment