Secure and Usable Audio CAPTCHAs

To prevent abuses of Internet services (e.g., automated account creation for sending spam mail), CAPTCHAs (Completely Automated Public Turing Tests to Tell Computers and Humans Apart) are used to distinguish human users from computer programs. Therefore, CAPTCHAs should represent a task that is easy to solve for humans but difficult for current machines.

In addition to the better-known image CAPTCHAs, audio CAPTCHAs are usually provided to enable access for visually impaired people. Previous studies show that most audio CAPTCHAs, albeit hard to solve for humans, are vulnerable as they can be broken by means of machine learning techniques.

My research project aims at creating highly secure audio CAPTCHAs that are usable by a wide audience. This is achieved by exploiting human capabilities that current machines are still lacking.

Time-frequency representation of a commercial audio CAPTCHA.
Time-frequency representation of a commercial audio CAPTCHA.