Us-r digital fingerprint stamp out scams for voice telephony

United States Georgia Institute of technology (Georgia Tech), the researchers said they define a hidden in speech signal in a digital fingerprint that can be used to prevent fraud.

When a caller is believed to be reliable, such as the credit card issuing bank — the phone asks you to provide your account password process seems quite reasonable, but unfortunately, the criminal group is actually very easy to spoof caller ID, and leverage and phishing similar techniques to deceive the public.

Leading information security center, Georgia Institute of technology (Information Security Center, GTISC) Mustaque Ahamad said of the Professor, the research and development of speech recognition technology that can apply a variety of phones, the initiative to 100% accuracy for caller identification.

Joint participation in the study, Dr. Vijay Balasubramaniyan, unlike untraceable electronic mail, voice signal will leave clues, revealing a pass calls must pass through a network patterns, such as advanced VOIP, wireless, mobile or fixed line of old.

The present call is very difficult to trace, as will every time through the network gateway is repeated series a few yards

Newly developed technology and can not be stated more voice location or IP address, but based on past call records to identify trusted callers, and calls for that phone and voucher does not comply with warn.

The system can use the phone to operate independently and are not required to pass the phone manufacturer or telecom operators to meet; this technology just to listen to the voice of the embedded "signature" to trace out the researchers of the so-called "last transmission sources (call's provenance)".

"The voice signal is essentially the crossing of some information embedded in the network, so we can from words to discriminatory conditions words;" Balasubramaniyan said: "for example, when a voice packet loss on the Internet, but the ear becomes aware of it.

"By collecting these subtle speech cues, including packets dropped the sound, the researchers developed a is called the Pindrop algorithm, can have from you all to answer the calls which identify each phone's unique digital signature.

Balasubramaniyan pointed out that, as long as there have been an incoming call, after the Pindrop algorithm will be able to 90% accuracy distinguish electricity; and because the system has the ability to learn continuously, in two-pass calls later, its accuracy can reach 96%.

Researchers and claims that there have been five-the same calls, the algorithm is able to 100% accuracy resolution out of power.

Researchers are also trying to expand the capacity of the Pindrop algorithm, for its accession to or no QSO records of cases resolved by the country's electricity.

It is understood that, at present, the research team had to distinguish from Australia, India, United Arab Emirates, United Kingdom and France, and call of different.