About

Musical CAPTCHA is a new approach to CAPTCHA, asking users to rate musical phrases tagged with human perception inspired labels.

What Is Captcha?

CAPTCHA (Completely Automated Public Turing Test to Tell Computers and Humans Apart) is a security mechanism to protect online systems from malicious attacks. The ideal CAPTCHA system should be easily solved by human participants, but not possible by a computer program [1]. The security of CAPTCHA is crucial to its design, however this often comes at the cost of usability and enjoyability, with users constantly being frustrated by their use [2].

A recent classification system identified 10 types of CAPTCHA: Text-based, Image-based, Audio-based, Video-based, Game-based, Slider-based, Behavior-based, Sensor-based, and CAPTCHAs for liveliness detection in authentication methods [3]. Considering existing audio CAPTCHA, the vast majority rely on speech, augmented with noise [4]. Audio CAPTCHA are currently primarily designed for those who are visually impaired, although common accuracy by users is only 50% [5]. The security of these systems is also lower than other CAPTCHA, with some common audio CAPTCHA schemes defeated by a bot 95% of the time [6].

CAPTCHA research is consistently developing, aiming to address six primary issues [3,7-9].

• security and ability to beat attacks
• usability by human users
• enjoyability; or at least as enjoyable as possible
• cross platform; between computer and mobile devices
• accessibility; use-able by as many as possible
• privacy; sharing as little use data as possible

Our Concept

We built our approach to leverage advances in music and artificial intelligence while addressing the previous six issues. Our approach combines generation of raw audio with natural language generation, to create non-semantic audio. We believe musical CAPTCHA can be an effective means for the non-disabled as an alternate form of CAPTCHA. as well as for visually impaired users.

Features of our Approach

• security
  • non-semantic audio receives significantly less attention than text, images or speech, with new approaches for classification requiring custom domain knowledge research
  • we use custom data sets that are specific for CAPTCHA, adding significant complexity to training an attack
• usability
  • in our past studies we have shown that users are able to accurately recognize our metrics 78% of the time
• enjoyability
  • in our past studies we have shown that 86% of users prefer our approach over traditional CAPTCHA
• cross platform
  • musical CAPTCHA can work with any device capable of outputting sound, no requirement for specific devices sensors
• accessibility
  • we aim to be accessible to as many users as possible, with options for vision-impaired
• privacy
  • we only require response to a sound, with no need for tracking human movements in any way.

How We Do It

We combine generation of raw music audio files, with language features passed on music, timbre (how it sounds) and emotion. Key features of our approach include:

• custom datasets, designed with human and musician tagged labels
• custom neural networks, leveraging state of the art approaches
• incorporation of musical perception driven processes

Citations

[1] L. v. Ahn, M. Blum, N. J. Hopper, and J. Langford, “Captcha: Using hard ai problems for security,” in International conference on the theory and applications of cryptographic techniques, pp. 294–311, Springer, 2003.

[2] J. Yan and A. S. El Ahmad, “Usability of captchas or usability issues in captcha design,” in Proceedings of the 4th symposium on Usable privacy and security, pp. 44–52, 2008.

[3] M. Guerar, L. Verderame, M. Migliardi, F. Palmieri, and A. Merlo, “Gotta captcha’em all: a survey of 20 years of the human-or-computer dilemma,” ACM Computing Surveys (CSUR), vol. 54, no. 9, pp. 1–33, 2021.

[4] S. Kulkarni and H. Fadewar, “Audio captcha techniques: A review,” in Proceedings of the Second International Conference on Computational Intelligence and Informatics, pp. 359–368, Springer, 2018.

[5] M. Alnfiai, “A novel design of audio captcha for visually impaired users,” International Journal of Communication Networks and Information Security, vol. 12, no. 2, pp. 168–179, 2020.

[6] K. Bock, D. Patel, G. Hughey, and D. Levin, “{unCaptcha}: A {Low-Resource} defeat of {reCaptcha’s} audio challenge,” in 11th USENIX Workshop on Offensive Technologies (WOOT 17), 2017.

[7] G. Ananth Kumar and G. Rishav, “Captcha techniques of secure web authentication: A survey,” in Communication Software and Networks, pp. 97–117, Springer, 2021.

[8] Y. Zhang, H. Gao, G. Pei, S. Luo, G. Chang, and N. Cheng, “A survey of research on captcha designing and breaking techniques,” in 2019 18th IEEE International Confer- ence On Trust, Security And Privacy In Computing And Communications/13th IEEE International Conference On Big Data Science And Engineering (TrustCom/BigDataSE), pp. 75–84, IEEE, 2019.

[9] M. Kumar, M. Jindal, and M. Kumar, “A systematic survey on captcha recognition: types, creation and breaking techniques,” Archives of Computational Methods in Engineering, vol. 29, no. 2, pp. 1107–1136, 2022.