Biometrics and Artificial Intelligence: Attacks and Challenges

L. Hernández-Álvarez, L. González-Manzano, J.M. Fuentes and L. Hernández Encinas
pp. 213–240 in the book “Breakthroughs in Digital Biometrics and Forensics”, Springer, 2022

In recent years, the development and design of reliable, secure, and usable user authentication systems have become increasingly important. One of the main reasons is the implementation and widespread use of online services, such as bank operations, online shopping, or access to personal and professional information. Since this information is sensitive and must be protected against non-authorized subjects, user authentication protocols are commonly applied. Artificial Intelligence (AI) algorithms and Biometrics are two tools usually combined to construct these protocols. While the first one allows extracting patterns and structures hidden in the data, the second one represents the characteristics that define how each person is. Therefore, these tools together can represent the manner in which a specific subject behaves and differentiate it from the rest of the users. In this chapter, we review the critical aspects of AI and Biometrics regarding user authentication protocols, so that the reader can get a general understanding of their combination. These aspects include the principal characteristics of both areas and how they can be joined in order to design reliable user authentication protocols. However, it should be considered that both fields, AI and Biometrics, present some limitations and risks that are inherent to them. Therefore, we also explain the potential attacks to which a user authentication system based on these tools is exposed. Finally, the principal limitations that are still unsolved are described.

Acknowledgements. This work was supported in part by the Spanish State Research Agency (AEI) of the Ministry of Science and Innovation (MCIN), project P2QProMeTe (PID2020-112586RB-I00/AEI/10.13039/501100011033); Project ODIO/COW (PID2019-111429RB-C21) and by Comunidad de Madrid (Spain) through Project CYNAMON (P2018/TCS-4566-CM), all of them co-funded by the European Regional Development Fund (ESF and ERDF, EU); in part by ORACLE Project (PCI2020-120691-2), funded by MCIN/AEI/10.13039/501100011033, and European Union “NextGenerationEU/PRTR”; in part by Comunidad de Madrid (Spain) with Universidad Carlos III de Madrid (Spain) under Grant CAVTIONS-CM-UC3M; and by the Madrid Government (Comunidad de Madrid, Spain) under the Multiannual Agreement with UC3M (“Fostering Young Doctors Research,” DEPROFAKE-CM-UC3M), and in the context of the V PRICIT (Research and Technological Innovation Regional Programme). L.H.A. would like to thank CSIC Project CASDiM for its support.