Telephone banking system recognizes your voice

RSA Security has introduced a product that adds voice as a way for automated telephone banking services to identify users.

The EMC subsidiary said Tuesday that the addition of voice biometrics to its Adaptive Authentication package marks an industry first. The product combines Vocent technology, acquired through its takeover of PassMark, with a voiceprint engine from speech recognition specialist Nuance Communications, RSA said in a statement.

The product is designed to help fight telephone banking fraud, the company said. "As we are strengthening security for the Web channel, phone banking is effectively becoming the next big target," Christopher Young, an RSA vice president, said in the statement.

The product generates a risk score by looking at the voiceprint as well as other parameters, such as the phone number and user behavior profiles, RSA said. Low-risk transactions proceed uninterrupted, while transactions with a high-risk score are verified with an additional layer of security, such as secret questions, it said.

Current phone authentication techniques lack security, as they are typically conducted semi-manually, and are susceptible to social engineering attacks, RSA said. Crooks are learning to take advantage of that, it added.

RSA's security products for the Web are used by 35 of the top 100 U.S. financial institutions and some large banks in Europe, the company said. The telephone banking security product is slated to be put into use at several banks in the first quarter of next year.

[Ngallendou]

Recorded voice lies

SCENE ONE
Crook (Gun to head of account holder): "Speak these words into to this microphone, or I will blow your head off! Name, account name, password, security answer, letters of the alphabet and numerals zero through nine"
Account holder (Do so.)
Crook (To partner): "Alright, phone up the bank and play these recordings in response to each query."
SCENE 2
Crook: "We have these recordings of the account holder's voice. What can you do with these?"
Partner: "Easy. We shall just copy out words and phrases, then concantinate them into responses to play over the telephone to the bank."

[Leonardo_da_Vinci]

Not really

The voice variability makes that if you use a gun, words from the user will flow with such a terrible feeling of desperation that they will not match the biometric model -created in calm- for sure. It is almost impossible because voice recognition is very influenced by extreme emotions. Sorry, but no way with weapons or threats!