Face spoofing detection via ensemble of classifiers toward low-power devices
- PDF / 1,675,198 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 73 Downloads / 208 Views
ORIGINAL ARTICLE
Face spoofing detection via ensemble of classifiers toward low‑power devices Rafael Henrique Vareto1 · William Robson Schwartz1 Received: 9 February 2020 / Accepted: 29 October 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Facial biometrics tend to be spontaneous, instinctive and less human intrusive. It is regularly employed in the authentication of authorized users and personnel to protect data from violation attacks. A face spoofing attack usually comprises the illegal attempt to access valuable undisclosed information as a trespasser attempts to impersonate an individual holding desirable authentication clearance. In search of such violations, many investigators have devoted their efforts to studying either visual liveness detection or patterns generated during media recapture as predominant indicators to block spoofing violations. This work contemplates low-power devices through the aggregation of Fourier transforms, different classification methods and handcrafted descriptors to estimate whether face samples correspond to falsification attacks. To the best of our knowledge, the proposed method consists of low computational cost and is one of the few methods associating features derived from both spatial and frequency image domains. We conduct experiments on recent and well-known datasets under same and cross-database settings with artificial neural networks, support vector machines and partial least squares ensembles. Results show that although our methodology is geared for resource-limited single-board computers, it can produce significant results, outperforming state-of-the-art approaches. Keywords Face spoofing · Liveness detection · Fourier transform · Machine learning · Biometrics
1 Introduction Biometrics is the science of automatically identifying individuals based on their physiological or behavioral characteristics, ranging from face and fingerprint to iris and voice. Despite the significant progress of biometric authentication techniques in the past years, experts declare that novel technologies are constantly exposed to malicious authentication attacks and can be susceptible to emerging high-quality fraudulent mechanisms [25]. The term spoofing, also known as copy and presentation attack, represents a serious threat to any biometric system. It eventuates when a criminal manipulates fraudulent data * Rafael Henrique Vareto [email protected] http://smartsenselab.dcc.ufmg.br/en William Robson Schwartz [email protected] 1
Smart Sense Laboratory, Department of Computer Science, Universidade Federal de Minas Gerais, Belo Horizonte, Minas Gerais, Brazil
to circumvent the security procedure and gain unauthorized access. More precisely, the attack occurs when an interloper attempts to impersonate someone who carries a desirable authentication clearance. As a countermeasure to presentation attacks, several researchers dedicate their time and efforts inspecting patterns generated during media recapture as well as building new databases as an at
Data Loading...
