Scholarlink

<p>Biometric-based recognition systems have gained significant attention in recent years due to their widespread deployment in security-critical applications. Face recognition systems which use facial biometrics, have become the most popular identification method after the COVID-19 pandemic. The security systems that use these technologies face major risks because hackers can exploit their weaknesses through presentation attacks, which people call face spoofing attacks. The field of face anti-spoofing (FAS) has developed multiple solutions which address these existing problems. The paper offers an extensive review that presents current advanced face anti-spoofing techniques while emphasizing deep learning-based methods. The research presents a comparative study that identifies how different methods perform through their strengths and weaknesses and their ability to develop new skills. In this study we present a new classification system that organizes FAS methods according to their attack scenarios and ability to generalize across different domains and their implementation challenges. The paper presents an extensive evaluation of available datasets and assessment methods used in public research. The research study presents essential research deficiencies together with upcoming research fields which require development of effective anti-spoofing methods that maintain performance in practical applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep learning based face anti spoofing mechanisms a comprehensive review of state of the art methods taxonomy and future trends

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月8日

查看全文
  • Syed Zoofa Rufai,
  • Saimul Bashir,
  • Faisal Firdous,
  • Suman Saurabh Sarkar

摘要

Biometric-based recognition systems have gained significant attention in recent years due to their widespread deployment in security-critical applications. Face recognition systems which use facial biometrics, have become the most popular identification method after the COVID-19 pandemic. The security systems that use these technologies face major risks because hackers can exploit their weaknesses through presentation attacks, which people call face spoofing attacks. The field of face anti-spoofing (FAS) has developed multiple solutions which address these existing problems. The paper offers an extensive review that presents current advanced face anti-spoofing techniques while emphasizing deep learning-based methods. The research presents a comparative study that identifies how different methods perform through their strengths and weaknesses and their ability to develop new skills. In this study we present a new classification system that organizes FAS methods according to their attack scenarios and ability to generalize across different domains and their implementation challenges. The paper presents an extensive evaluation of available datasets and assessment methods used in public research. The research study presents essential research deficiencies together with upcoming research fields which require development of effective anti-spoofing methods that maintain performance in practical applications.